FlippyDora commited on
Commit
7fb613d
·
verified ·
1 Parent(s): 4cf2485

Add runs (part 27)

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +2 -0
  2. scenes/isaac/rcd_lounge_4/20261114T065436Z/generation.jsonl +3 -0
  3. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/anchor_marked.png +3 -0
  4. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/anchor_reference.png +3 -0
  5. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/generation_spec.json +66 -0
  6. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/prompt.txt +143 -0
  7. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/scale_anchor.json +26 -0
  8. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/sdk_smoke.py +49 -0
  9. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/self_check.py +228 -0
  10. scenes/isaac/rcd_lounge_4/20261114T065436Z/input/video.mp4 +3 -0
  11. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_fix.py +21 -0
  12. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_fix2.py +24 -0
  13. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_sheet_all.png +3 -0
  14. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/build_scene.py +510 -0
  15. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/objects.json +138 -0
  16. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/render_preview.py +144 -0
  17. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/scene.usda +0 -0
  18. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/textures/carpet.png +3 -0
  19. scenes/isaac/rcd_lounge_4/20261114T065436Z/output/textures/quilt.png +3 -0
  20. scenes/isaac/rcd_lounge_4/20261114T065436Z/run.json +139 -0
  21. scenes/isaac/rcd_lounge_4/20261114T065436Z/submitted_prompt.txt +143 -0
  22. scenes/isaac/rcd_lounge_4/20261114T083703Z/evaluator_protocol.json +0 -0
  23. scenes/isaac/rcd_lounge_4/20261114T083703Z/generation.jsonl +0 -0
  24. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/anchor_marked.png +3 -0
  25. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/anchor_reference.png +3 -0
  26. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/generation_spec.json +66 -0
  27. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/prompt.txt +143 -0
  28. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/scale_anchor.json +26 -0
  29. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/sdk_smoke.py +49 -0
  30. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/self_check.py +228 -0
  31. scenes/isaac/rcd_lounge_4/20261114T083703Z/input/video.mp4 +3 -0
  32. scenes/isaac/rcd_lounge_4/20261114T083703Z/run.json +139 -0
  33. scenes/isaac/rcd_lounge_4/20261114T083703Z/submitted_prompt.txt +143 -0
  34. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/evaluator_protocol.json +0 -0
  35. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/generation.jsonl +3 -0
  36. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/anchor_marked.png +3 -0
  37. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/anchor_reference.png +3 -0
  38. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/generation_spec.json +66 -0
  39. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/prompt.txt +143 -0
  40. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/scale_anchor.json +26 -0
  41. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/sdk_smoke.py +49 -0
  42. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/self_check.py +228 -0
  43. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/video.mp4 +3 -0
  44. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/build_scene.py +551 -0
  45. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/scene.usda +2 -0
  46. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/carpet.png +3 -0
  47. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/dots.png +3 -0
  48. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/wood.png +3 -0
  49. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/run.json +139 -0
  50. scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/submitted_prompt.txt +143 -0
.gitattributes CHANGED
@@ -5904,3 +5904,5 @@ scenes/isaac/44358647/20261114T061621Z/generation.jsonl filter=lfs diff=lfs merg
5904
  scenes/isaac/45662970/20261114T064610Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5905
  scenes/isaac/47115300/20261114T062346Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5906
  scenes/isaac/47429874/20261114T062029Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
 
 
 
5904
  scenes/isaac/45662970/20261114T064610Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5905
  scenes/isaac/47115300/20261114T062346Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5906
  scenes/isaac/47429874/20261114T062029Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5907
+ scenes/isaac/rcd_lounge_4/20261114T065436Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
5908
+ scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/generation.jsonl filter=lfs diff=lfs merge=lfs -text
scenes/isaac/rcd_lounge_4/20261114T065436Z/generation.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4660a899367db6a8b639958a8e62e6d36992b34d21e71212493d4e2e1dd2cd62
3
+ size 19280962
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/anchor_marked.png ADDED

Git LFS Details

  • SHA256: fc219ca78ee0730aaefac3d910420664fcf83c83747d04c6a8dfc97a3ec85405
  • Pointer size: 131 Bytes
  • Size of remote file: 272 kB
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/anchor_reference.png ADDED

Git LFS Details

  • SHA256: 02d769e29bd08628f82ebb850ed062b88897e9225aabc5134b12f5419dd2d7f8
  • Pointer size: 131 Bytes
  • Size of remote file: 273 kB
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/generation_spec.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "contract": "scenes-isaac",
3
+ "engine": "Isaac Sim",
4
+ "task": "Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.",
5
+ "inputs": [
6
+ "video.mp4",
7
+ "scale_anchor.json",
8
+ "anchor_reference.png",
9
+ "anchor_marked.png",
10
+ "self_check.py"
11
+ ],
12
+ "prompt": "prompt.txt",
13
+ "units": "metres, kilograms, seconds",
14
+ "up_axis": "+Z",
15
+ "entrypoint": "scene.usda",
16
+ "budget": {
17
+ "wall_clock_seconds": 1800,
18
+ "scope": "per_scene",
19
+ "starts_at": "agent_input_access",
20
+ "includes_all_tool_and_validation_time": true,
21
+ "deadline_action": "stop and submit latest valid loadable output",
22
+ "enforcement": "Docker generation runner terminates the container at the wall-clock deadline"
23
+ },
24
+ "requirement": "Sim-ready: ready for interactive robot learning, with no manual repair.",
25
+ "sim_ready": {
26
+ "loads_headless_offline": true,
27
+ "resets_to_rest": true,
28
+ "collision_matches_visible_geometry": true,
29
+ "rigid_things": "rigid bodies; free objects movable, articulated parts on limited joints",
30
+ "soft_things": [
31
+ "cloth",
32
+ "bedding",
33
+ "curtains",
34
+ "rugs",
35
+ "throws",
36
+ "cushions",
37
+ "pillows"
38
+ ],
39
+ "soft_mechanism": "native deformable",
40
+ "stable_under_sustained_interaction": true
41
+ },
42
+ "visual_objective": "Maximize visual quality, realism and fidelity to the source video, keeping every sim-ready requirement.",
43
+ "refinement": {
44
+ "minimum_completed_rounds": 3,
45
+ "minimum_reference_views": 6,
46
+ "minimum_inspected_video_frames": 12,
47
+ "final_validation_seconds": 90,
48
+ "outside_data_allowed": false
49
+ },
50
+ "label": "free text saying plainly what the object is; one entry per physical instance",
51
+ "interaction_points": [
52
+ "handle",
53
+ "grasp",
54
+ "spout",
55
+ "material_corners"
56
+ ],
57
+ "construction_rules": {
58
+ "max_initial_penetration_m": 0.005,
59
+ "max_penetration_duration_s": 0.02,
60
+ "max_passive_speed_mps": 5.0,
61
+ "articulated_parts_clear_of_carcass": true,
62
+ "swing_paths_clear": true,
63
+ "timestep": "agent's choice"
64
+ },
65
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check"
66
+ }
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 30 minutes (1,800 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/scale_anchor.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "object_id": "scale_anchor",
3
+ "description": "The couch (2) marked in red in anchor_marked.png.",
4
+ "reference_image": "anchor_reference.png",
5
+ "marked_image": "anchor_marked.png",
6
+ "region_xyxy_pixels": [
7
+ 56.39891520732052,
8
+ 6.302853824313528,
9
+ 324.97904181777216,
10
+ 223.66539350903474
11
+ ],
12
+ "region_type": "projected_3d_box_envelope_not_segmentation",
13
+ "size_local_xyz_m": [
14
+ 1.256,
15
+ 0.838,
16
+ 0.859
17
+ ],
18
+ "axis_description": [
19
+ "horizontal length",
20
+ "horizontal width",
21
+ "vertical height"
22
+ ],
23
+ "measurement_type": "full extents of an oriented box around the object",
24
+ "uncertainty_m": null,
25
+ "uncertainty_note": "No uncertainty is supplied for these dimensions."
26
+ }
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/sdk_smoke.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Verify Isaac Sim headless RTX rendering and native rigid-body stepping."""
2
+ from isaacsim import SimulationApp
3
+ app = SimulationApp({'headless':True,'renderer':'PathTracing','samples_per_pixel_per_frame':32,
4
+ 'width':512,'height':384,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
5
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
6
+ import json
7
+ from pathlib import Path
8
+ import numpy as np
9
+ from PIL import Image
10
+ import omni.usd
11
+ import omni.replicator.core as rep
12
+ rep.orchestrator.set_capture_on_play(False)
13
+ from pxr import UsdGeom,UsdLux,UsdShade,UsdPhysics,Gf,Sdf
14
+ stage=omni.usd.get_context().get_stage()
15
+ UsdGeom.SetStageUpAxis(stage,UsdGeom.Tokens.z);UsdGeom.SetStageMetersPerUnit(stage,1.)
16
+ UsdGeom.Xform.Define(stage,'/World');stage.SetDefaultPrim(stage.GetPrimAtPath('/World'))
17
+ def box(path,pos,scale,color,dynamic=False):
18
+ prim=UsdGeom.Cube.Define(stage,path);prim.CreateSizeAttr(1.)
19
+ xf=UsdGeom.Xformable(prim);xf.AddTranslateOp().Set(Gf.Vec3d(*pos));xf.AddScaleOp().Set(Gf.Vec3f(*scale))
20
+ UsdPhysics.CollisionAPI.Apply(prim.GetPrim())
21
+ if dynamic:
22
+ UsdPhysics.RigidBodyAPI.Apply(prim.GetPrim());UsdPhysics.MassAPI.Apply(prim.GetPrim()).CreateMassAttr(1.)
23
+ mat=UsdShade.Material.Define(stage,path+'_Material');shader=UsdShade.Shader.Define(stage,path+'_Material/Shader');shader.CreateIdAttr('UsdPreviewSurface')
24
+ shader.CreateInput('diffuseColor',Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*color));shader.CreateInput('roughness',Sdf.ValueTypeNames.Float).Set(.28)
25
+ mat.CreateSurfaceOutput().ConnectToSource(shader.ConnectableAPI(),'surface');UsdShade.MaterialBindingAPI.Apply(prim.GetPrim()).Bind(mat)
26
+ return prim
27
+ box('/World/Floor',(0,0,-.05),(5,5,.1),(.6,.6,.6));box('/World/Cube',(0,0,1.),(.3,.3,.3),(.08,.3,.8),True)
28
+ light=UsdLux.DomeLight.Define(stage,'/World/Dome');light.CreateIntensityAttr(500.)
29
+ light=UsdLux.RectLight.Define(stage,'/World/Key');light.CreateIntensityAttr(2500.);light.CreateWidthAttr(3.);light.CreateHeightAttr(3.);UsdGeom.Xformable(light).AddTranslateOp().Set(Gf.Vec3d(0,0,3))
30
+ scene=UsdPhysics.Scene.Define(stage,'/World/PhysicsScene');scene.CreateGravityDirectionAttr(Gf.Vec3f(0,0,-1));scene.CreateGravityMagnitudeAttr(9.81)
31
+ camera=rep.create.camera(position=(2.4,-3,2),look_at=(0,0,.3));product=rep.create.render_product(camera,(512,384));rgb=rep.AnnotatorRegistry.get_annotator('rgb');rgb.attach([product])
32
+ from isaacsim.core.api import World
33
+ world=World(physics_dt=1/60,rendering_dt=1/60,stage_units_in_meters=1.)
34
+ world.reset()
35
+ for _ in range(30):world.step(render=True)
36
+ for attempt in range(30):
37
+ rep.orchestrator.step(rt_subframes=8)
38
+ app.update()
39
+ image=np.asarray(rgb.get_data())
40
+ print('CAPTURE_ATTEMPT',attempt,'SHAPE',image.shape,flush=True)
41
+ if image.ndim==3 and image.shape[0]>0:break
42
+ print('RGB_SHAPE',image.shape,flush=True)
43
+ assert image.ndim==3,'RTX annotator did not return a frame'
44
+ position=UsdGeom.Xformable(stage.GetPrimAtPath('/World/Cube')).ComputeLocalToWorldTransform(0).ExtractTranslation()
45
+ print('CUBE_POSITION',list(position),flush=True)
46
+ out=Path('/work/output');out.mkdir(exist_ok=True)
47
+ Image.fromarray(image[:,:,:3]).save(out/'smoke.png');stage.GetRootLayer().Export(str(out/'sdk_example.usda'))
48
+ (out/'smoke.json').write_text(json.dumps({'renderer':'Isaac Sim RTX PathTracing','shape':list(image.shape),'cube_final_position':list(position),'initial_height_m':1.,'success':True}))
49
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/self_check.py ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Passive physics of a USD scene in Isaac Sim, against the scene requirements (no penetration beyond 5 mm for more than
2
+ 0.02 s, no body faster than 5 m/s, finite state).
3
+
4
+ Runs inside the Isaac Sim container. Reads --scene (default /work/scene, read-only) and the optional
5
+ --out/request.json (object ids to drop); writes --out/physics.json (default /work/physics). Without a request it is
6
+ the self-check: no drops, plus a check that every object has a non-empty label.
7
+
8
+ - Speed: a 3 s passive rollout with no rendering and no state reads, timed around world.step only
9
+ (simulated seconds / wall seconds).
10
+ - Passive rollout: 3 s from the authored state. Penetration comes from the PhysX tensor contact view
11
+ (GPU contact reports return nothing under GPU dynamics); speed and poses from the RigidPrim view.
12
+ - Drops: the requested objects with free (unjointed) dynamic bodies, raised 0.05 m
13
+ and released for 2 s, each from a fresh reset.
14
+ The limits are the same as for MuJoCo scenes.
15
+ """
16
+ from isaacsim import SimulationApp
17
+ app=SimulationApp({'headless':True,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
18
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
19
+ import argparse
20
+ import json
21
+ from pathlib import Path
22
+ import time
23
+ import carb
24
+ import numpy as np
25
+ import omni.usd
26
+ from pxr import UsdPhysics,PhysxSchema
27
+
28
+ LIMIT_M,LIMIT_S,SPEED=.005,.02,5.
29
+ parser=argparse.ArgumentParser();parser.add_argument('--scene',default='/work/scene');parser.add_argument('--out',default='/work/physics')
30
+ args,_=parser.parse_known_args()
31
+ out=Path(args.out);out.mkdir(parents=True,exist_ok=True)
32
+ request=json.loads((out/'request.json').read_text()) if (out/'request.json').is_file() else {'object_ids':[]}
33
+ result={'status':'failed','engine':'Isaac Sim PhysX (GPU dynamics)','no_vlm':True,'provisional':True,
34
+ 'thresholds':{'penetration_m':LIMIT_M,'penetration_duration_s':LIMIT_S,'linear_speed_mps':SPEED},
35
+ 'thresholds_calibrated_across_engines':False}
36
+
37
+
38
+ def contact_filters(stage,paths):
39
+ """GPU contact views filter on rigid bodies or standalone colliders only, so a collider inside a
40
+ rigid body is replaced by that body (unsupported filters are dropped with only a warning)."""
41
+ out=[]
42
+ for path in paths:
43
+ if any('deformable' in str(a).lower() for a in stage.GetPrimAtPath(path).GetAppliedSchemas()):
44
+ continue # deformable meshes cannot be contact filters; their contacts are not measured
45
+ prim=stage.GetPrimAtPath(path);owner=None
46
+ while prim and not prim.IsPseudoRoot():
47
+ if prim.HasAPI(UsdPhysics.RigidBodyAPI):owner=str(prim.GetPath())
48
+ prim=prim.GetParent()
49
+ target=owner or path
50
+ if target not in out:out.append(target)
51
+ return out
52
+
53
+
54
+ def array(value):
55
+ return np.asarray(value.cpu() if hasattr(value,'cpu') else value,dtype=float)
56
+
57
+
58
+ try:
59
+ root=Path(args.scene)
60
+ scene=next(root/n for n in ['scene.usda','scene.usd'] if (root/n).is_file())
61
+ omni.usd.get_context().open_stage(str(scene));stage=omni.usd.get_context().get_stage()
62
+ for _ in range(8):app.update()
63
+
64
+ def dynamic(prim):
65
+ api=UsdPhysics.RigidBodyAPI(prim)
66
+ if api.GetRigidBodyEnabledAttr().Get() not in (None,True) or api.GetKinematicEnabledAttr().Get():return False
67
+ parent=prim.GetParent()
68
+ while parent and not parent.IsPseudoRoot():
69
+ if parent.HasAPI(UsdPhysics.RigidBodyAPI):return False
70
+ parent=parent.GetParent()
71
+ return True
72
+ rigid=[p for p in stage.Traverse() if p.HasAPI(UsdPhysics.RigidBodyAPI)]
73
+ bodies=[p for p in rigid if dynamic(p)]
74
+ paths=[str(p.GetPath()) for p in bodies]
75
+ colliders=[str(p.GetPath()) for p in stage.Traverse() if p.HasAPI(UsdPhysics.CollisionAPI)]
76
+ jointed=set()
77
+ for prim in stage.Traverse():
78
+ if prim.IsA(UsdPhysics.Joint):
79
+ joint=UsdPhysics.Joint(prim)
80
+ for rel in (joint.GetBody0Rel(),joint.GetBody1Rel()):
81
+ jointed.update(str(t) for t in rel.GetTargets())
82
+ for prim in bodies:PhysxSchema.PhysxContactReportAPI.Apply(prim).CreateThresholdAttr().Set(0.)
83
+ scenes=[p for p in stage.Traverse() if p.IsA(UsdPhysics.Scene)]
84
+ authored=PhysxSchema.PhysxSceneAPI(scenes[0]).GetTimeStepsPerSecondAttr().Get() if scenes else None
85
+ rate=float(authored or 60)
86
+ result.update(dynamic_bodies=len(paths),colliders=len(colliders),
87
+ excluded_bodies=[str(p.GetPath()) for p in rigid if not dynamic(p)],
88
+ timestep_s=1/rate,timestep_source='authored PhysxSceneAPI timeStepsPerSecond' if authored else 'default 60 Hz')
89
+ if not paths:raise ValueError('No dynamic rigid bodies to simulate')
90
+ filters=contact_filters(stage,colliders)
91
+ result['contact_filters']=len(filters)
92
+
93
+ carb.settings.get_settings().set_bool('/physics/disableContactProcessing',False)
94
+ from isaacsim.core.api import World
95
+ from isaacsim.core.prims import RigidPrim
96
+ from isaacsim.core.simulation_manager import SimulationManager
97
+ world=World(physics_dt=1/rate,rendering_dt=1/rate,stage_units_in_meters=1.,
98
+ physics_prim_path=str(scenes[0].GetPath()) if scenes else '/World/AddedPhysicsScene',
99
+ set_defaults=False,backend='torch',device='cuda:0')
100
+ context=world.get_physics_context()
101
+ context.enable_gpu_dynamics(True);context.set_broadphase_type('GPU')
102
+ view=RigidPrim(paths,name='scene_bodies');world.scene.add(view)
103
+ steps=lambda seconds:int(np.ceil(seconds*rate))
104
+
105
+ def measure_rtf(dt,repeats=3):
106
+ """Simulated seconds / wall seconds of world.step(render=False) for an unread passive rollout from the authored
107
+ state, median over `repeats` after 10 warm-up steps."""
108
+ world.set_simulation_dt(physics_dt=dt,rendering_dt=dt)
109
+ world.reset()
110
+ for _ in range(10):world.step(render=False)
111
+ count=int(np.ceil(3/dt));samples=[];finite=True
112
+ for _ in range(repeats):
113
+ world.reset();start=time.perf_counter()
114
+ for _ in range(count):world.step(render=False)
115
+ samples.append(count*dt/(time.perf_counter()-start))
116
+ positions,_=view.get_world_poses()
117
+ finite&=bool(np.isfinite(array(positions)).all() and np.isfinite(array(view.get_velocities())).all())
118
+ return {'rtf':float(np.median(samples)),'rtf_samples':samples,'timestep_s':dt,'simulated_seconds':count*dt,
119
+ 'steps':count,'repeats':repeats,'finite':finite,
120
+ 'definition':'simulated seconds / wall seconds of world.step(render=False) only, median of repeats'}
121
+
122
+ result['realtime']={**measure_rtf(1/rate),'scope':'3 s passive rollout from the authored state, authored timestep'}
123
+ print('PHYSICS_STAGE realtime',result['realtime']['rtf'],flush=True)
124
+
125
+ def reset():
126
+ world.reset()
127
+ contacts=SimulationManager.get_physics_sim_view().create_rigid_contact_view(
128
+ paths,filter_patterns=[filters]*len(paths),max_contact_data_count=200000)
129
+ if contacts is None:raise RuntimeError('PhysX rigid contact view could not be created')
130
+ return contacts
131
+
132
+ def rollout(seconds,contacts,motion=None):
133
+ """One rollout; `motion` names the file the logged poses are written to."""
134
+ report={'finite':True,'peak_penetration_m':0.,'penetration_time_s':0.,'peak_linear_speed_mps':0.}
135
+ every=max(1,round(rate/10));frames=[]
136
+ for step in range(steps(seconds)):
137
+ world.step(render=False)
138
+ positions,orientations=view.get_world_poses();velocities=array(view.get_velocities())
139
+ if step%every==0:frames.append(((step+1)/rate,array(positions),array(orientations)))
140
+ report['finite']&=bool(np.isfinite(array(positions)).all() and np.isfinite(velocities).all())
141
+ _,_,_,separation,counts,starts=contacts.get_contact_data(dt=1/rate)
142
+ total=int(array(counts).sum())
143
+ if 'initial_pairs' not in report:
144
+ # Pairs deeper than the limit after the first step, named for the self-check.
145
+ pairs=[];sep=array(separation);cnt=array(counts).astype(int);beg=array(starts).astype(int)
146
+ for i,j in zip(*np.nonzero(cnt)):
147
+ depth=-float(sep[beg[i,j]:beg[i,j]+cnt[i,j]].min())
148
+ if depth>LIMIT_M:pairs.append({'depth_m':depth,'bodies':[paths[i],filters[j]]})
149
+ report['initial_pairs']=sorted(pairs,key=lambda p:-p['depth_m'])[:20]
150
+ deepest=max(0.,-float(array(separation)[:total].min())) if total else 0.
151
+ report['peak_penetration_m']=max(report['peak_penetration_m'],deepest)
152
+ if deepest>LIMIT_M:report['penetration_time_s']+=1/rate
153
+ report['peak_linear_speed_mps']=max(report['peak_linear_speed_mps'],float(np.linalg.norm(velocities[:,:3],axis=1).max()))
154
+ if not report['finite']:break
155
+ report['valid']=bool(report['finite'] and report['penetration_time_s']<=LIMIT_S and report['peak_linear_speed_mps']<=SPEED)
156
+ report['thresholds']=result['thresholds']
157
+ if motion and request.get('record'):
158
+ (out/'motion').mkdir(exist_ok=True)
159
+ np.savez_compressed(out/'motion'/motion,paths=np.array(paths),t=np.array([f[0] for f in frames]),
160
+ positions=np.stack([f[1] for f in frames]),orientations=np.stack([f[2] for f in frames]))
161
+ report['motion']=f'motion/{motion}.npz'
162
+ return report
163
+
164
+ contacts=reset()
165
+ start_positions=array(view.get_world_poses()[0])
166
+ masses=array(view.get_masses())
167
+ direction,magnitude=context.get_gravity()
168
+ gravity=np.asarray(direction,dtype=float)*float(magnitude)
169
+ passive=rollout(3,contacts,'passive')
170
+ print('PHYSICS_STAGE passive',passive,flush=True)
171
+ end_positions=array(view.get_world_poses()[0])
172
+ moved=np.linalg.norm(end_positions-start_positions,axis=1)
173
+
174
+ manifest=json.loads((root/'objects.json').read_text())
175
+ manifest=manifest['objects'] if isinstance(manifest,dict) else manifest
176
+ result['manifest_errors']=[f"{o.get('id')}: label is missing or empty" for o in manifest if not str(o.get('label') or '').strip()]
177
+ result['manifest_errors']+=[f"{o.get('id')}: prim_path does not resolve" for o in manifest if not stage.GetPrimAtPath(o.get('prim_path','') or '/')]
178
+ def owned(obj):
179
+ prefix=str(obj.get('prim_path','')).rstrip('/')
180
+ return [i for i,p in enumerate(paths) if prefix and (p==prefix or p.startswith(prefix+'/'))]
181
+ drift=[{'id':o['id'],'dynamic_bodies':len(owned(o)),'center_drift_m':float(moved[owned(o)].max()),'mass_kg':float(masses[owned(o)].sum())}
182
+ for o in manifest if owned(o)]
183
+
184
+ requested=set(request['object_ids'])
185
+ eligible=[o for o in manifest if o['id'] in requested]
186
+ drops=[]
187
+ for obj in eligible:
188
+ free=[i for i in owned(obj) if paths[i] not in jointed]
189
+ if not free:continue
190
+ contacts=reset()
191
+ positions,orientations=view.get_world_poses()
192
+ raised=positions.clone() if hasattr(positions,'clone') else positions.copy()
193
+ raised[free,2]+=.05
194
+ view.set_world_poses(positions=raised,orientations=orientations)
195
+ applied=float((array(view.get_world_poses()[0])[free,2]-array(positions)[free,2]).min())
196
+ drops.append({'id':obj['id'],'input':'raise free bodies 0.05 m and release','bodies':[paths[i] for i in free],
197
+ 'lift_applied_m':applied,**rollout(2,contacts,f'drop_{len(drops)}')})
198
+ print('PHYSICS_STAGE drop',obj['id'],applied,drops[-1]['peak_linear_speed_mps'],flush=True)
199
+
200
+ # The common 2 ms step last, so the rollouts above run at the authored step.
201
+ result['realtime_common_step']={**measure_rtf(.002),'scope':'3 s passive rollout from the authored state at the common 2 ms step'}
202
+ print('PHYSICS_STAGE realtime_common_step',result['realtime_common_step']['rtf'],flush=True)
203
+ gravity_ok=bool(np.allclose(gravity,[0,0,-9.81],atol=.1))
204
+ mass_ok=bool(np.isfinite(masses).all() and (masses>=0).all())
205
+ result.update(status='completed',passive=passive,gravity_m_s2=gravity.tolist(),gravity_valid=gravity_ok,
206
+ mass_valid=mass_ok,center_drift=drift,drop_tests=drops,eligible_native_objects=len(eligible),
207
+ largest_body_displacements=[{'prim':paths[i],'displacement_m':float(moved[i])} for i in np.argsort(-moved)[:5]],
208
+ valid=bool(passive['valid'] and gravity_ok and mass_ok),
209
+ drop_scope='the requested objects with free (unjointed) dynamic bodies, in manifest order',
210
+ contact_scope=('contacts between each dynamic body and every collider (static, kinematic and dynamic; colliders '
211
+ 'inside a rigid body are filtered by that body, as GPU contact views require); '
212
+ 'static-static pairs are not simulated, as in MuJoCo world-body geoms'),
213
+ limitations=['the limits are the MuJoCo ones, not calibrated for PhysX',
214
+ 'kinematic, disabled and nested rigid bodies are not traced (see excluded_bodies)',
215
+ 'drift is per dynamic body under the object prim, not the object bounding box',
216
+ 'contacts with deformable bodies are not measured (GPU contact views cannot filter them)'])
217
+ except Exception as error:
218
+ result['error']=f'{type(error).__name__}: {error}'
219
+ finally:
220
+ (out/'physics.json').write_text(json.dumps(result,indent=2)+'\n')
221
+ if 'passive' in result:
222
+ passive=result['passive']
223
+ print('PASSIVE CHECK','PASSED' if passive['valid'] and not result.get('manifest_errors') else 'FAILED',
224
+ f"- {len(passive.get('initial_pairs',[]))} contact pairs > 5 mm, {passive['penetration_time_s']:.3f} s above 5 mm,"
225
+ f" peak speed {passive['peak_linear_speed_mps']:.2f} m/s, {len(result.get('manifest_errors',[]))} manifest errors;"
226
+ f" details in {out/'physics.json'}",flush=True)
227
+ else:print('CHECK FAILED:',result.get('error'),flush=True)
228
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T065436Z/input/video.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:266549020e323df9619a00d353edef099d840ba582bfa732ccc7af90220e1fe4
3
+ size 16239519
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_fix.py ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ f="/work/output/build_scene.py"; s=open(f).read()
2
+ repls=[
3
+ ('orange_chair("orange_chair_1", (-2.6, 1.5), 25)','orange_chair("orange_chair_1", (-2.7, 1.7), 25)'),
4
+ ('orange_chair("orange_chair_2", (-1.85, 0.35), 55)','orange_chair("orange_chair_2", (-2.2, 0.3), 55)'),
5
+ ('rigid("lavender_armchair", (2.75, 2.3), 20.0','rigid("lavender_armchair", (2.9, 2.5), 20.0'),
6
+ ('sculptural("sculptural_chair_1", (1.45, 0.55)','sculptural("sculptural_chair_1", (0.9, 0.2)'),
7
+ ('sculptural("scale_anchor", (2.6, 0.1)','sculptural("scale_anchor", (2.6, 0.1)'),
8
+ ('sculptural("scale_anchor", (2.7, 0.4)','sculptural("scale_anchor", (2.6, 0.1)'),
9
+ ('rigid("white_appliance", (-3.05, 2.65), 9.0','rigid("white_appliance", (-3.1, 2.7), 9.0'),
10
+ ('rigid("blue_box", (1.7, -1.35), 2.0','rigid("blue_box", (1.6, -1.4), 2.0'),
11
+ ('rigid("water_pitcher", (3.2, -1.9), 0.7','rigid("water_pitcher", (3.3, -2.2), 0.7'),
12
+ ('AddTranslateOp().Set(Gf.Vec3d(3.4, 1.0, 0))','AddTranslateOp().Set(Gf.Vec3d(3.4, -1.0, 0))'),
13
+ ('AddTranslateOp().Set(Gf.Vec3d(3.4, 0.92, 0.30))','AddTranslateOp().Set(Gf.Vec3d(3.4, -1.08, 0.30))'),
14
+ ('deformable("pebble_ottoman_1", (0.55, -1.75, 0.22)','deformable("pebble_ottoman_1", (0.4, -1.8, 0.22)'),
15
+ ('deformable("pebble_ottoman_2", (-1.05, -2.0, 0.22)','deformable("pebble_ottoman_2", (-1.2, -2.2, 0.22)'),
16
+ ('deformable("accent_rug", (-2.4, -2.1, 0.012)','deformable("accent_rug", (-2.6, -2.3, 0.012)'),
17
+ ('revolute_door("glass_door", (-1.0, -3.44, 0.75), (-0.5, -3.44, 0.75)','revolute_door("glass_door", (-1.0, -3.38, 0.75), (-0.5, -3.38, 0.75)'),
18
+ ]
19
+ for a,b in repls:
20
+ if a in s: s=s.replace(a,b)
21
+ open(f,"w").write(s); print("fix done")
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_fix2.py ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ f="/work/output/build_scene.py"; s=open(f).read()
2
+ # sphere/cylinder colliders are oversized -> visual only; cubes are the colliders
3
+ s=s.replace(" if collision:\n UsdPhysics.CollisionAPI.Apply(g.GetPrim())",
4
+ " if collision and kind == \"cube\":\n UsdPhysics.CollisionAPI.Apply(g.GetPrim())")
5
+ # add a hidden box collision proxy inside each sculptural chair's rounded lobes
6
+ old=''' ("sphere", (0, -0.28, 0.58), (0.62, 0.34, 0.62), QUILT), # tall back
7
+ ("cube", (0, 0, 0.06), (1.15, 0.72, 0.12), PEBBL), # base skirt
8
+ ]'''
9
+ new=''' ("sphere", (0, -0.28, 0.58), (0.62, 0.34, 0.62), QUILT), # tall back
10
+ ("cube", (0, 0, 0.30), (1.02, 0.66, 0.52), PEBBL), # collision proxy
11
+ ("cube", (0, 0, 0.06), (1.15, 0.72, 0.12), PEBBL), # base skirt
12
+ ]'''
13
+ assert old in s, "sculptural block not found"; s=s.replace(old,new)
14
+ # pitcher body collision proxy (inside the cylinder)
15
+ old2='''rigid("water_pitcher", (3.3, -2.2), 0.7, [
16
+ ("cyl", (0, 0, 0.16), (0.16, 0.16, 0.32), solid((0.75, 0.8, 0.85), 0.35)),'''
17
+ new2='''rigid("water_pitcher", (3.3, -2.2), 0.7, [
18
+ ("cube", (0, 0, 0.16), (0.15, 0.15, 0.30), solid((0.75, 0.8, 0.85), 0.35)),
19
+ ("cyl", (0, 0, 0.16), (0.16, 0.16, 0.32), solid((0.75, 0.8, 0.85), 0.35)),'''
20
+ assert old2 in s, "pitcher block not found"; s=s.replace(old2,new2)
21
+ # move glass door clear of front wall
22
+ s=s.replace('revolute_door("glass_door", (-1.0, -3.38, 0.75), (-0.5, -3.38, 0.75)',
23
+ 'revolute_door("glass_door", (-1.0, -3.2, 0.75), (-0.5, -3.2, 0.75)')
24
+ open(f,"w").write(s); print("fix2 done")
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/_sheet_all.png ADDED

Git LFS Details

  • SHA256: 56796498473f98b2dd9c15bb58b786aab2ce5c27949a6ccaed0c5f549da3641c
  • Pointer size: 132 Bytes
  • Size of remote file: 4.66 MB
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/build_scene.py ADDED
@@ -0,0 +1,510 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Reproducible builder for the lounge/waiting-room reconstruction.
2
+
3
+ Run with: /isaac-sim/python.sh /work/output/build_scene.py
4
+ Produces a self-contained Isaac Sim / OpenUSD stage (Z-up, metres) with an
5
+ authored PhysX scene, rigid bodies, articulated joints, and native PhysX
6
+ volume-deformable soft items. All textures are generated locally.
7
+ """
8
+ import os, json, math, sys
9
+ from isaacsim import SimulationApp
10
+ app = SimulationApp({'headless': True, 'multi_gpu': False, 'active_gpu': 0,
11
+ 'physics_gpu': 0, 'enable_crashreporter': False,
12
+ 'limit_cpu_threads': 8})
13
+ import numpy as np
14
+ from PIL import Image, ImageDraw
15
+ import omni.usd
16
+ from pxr import Usd, UsdGeom, UsdShade, UsdPhysics, UsdLux, PhysxSchema, Sdf, Gf, Vt
17
+
18
+ OUT = "/work/output"
19
+ TEX = os.path.join(OUT, "textures")
20
+ os.makedirs(TEX, exist_ok=True)
21
+ USE_DEFORM = "--no-deform" not in sys.argv
22
+
23
+ # ---------------------------------------------------------------- textures
24
+ def carpet_texture():
25
+ N = 512; T = 4; cell = N // T
26
+ img = Image.new("RGB", (N, N)); px = img.load()
27
+ rng = np.random.default_rng(7)
28
+ base = [(150, 140, 126), (120, 124, 126), (163, 152, 136),
29
+ (108, 112, 118), (140, 133, 120), (96, 104, 112)]
30
+ for ty in range(T):
31
+ for tx in range(T):
32
+ b = np.array(base[(tx * 3 + ty * 5) % len(base)], float)
33
+ horiz = ((tx + ty) % 2 == 0)
34
+ for y in range(cell):
35
+ for x in range(cell):
36
+ gx, gy = tx * cell + x, ty * cell + y
37
+ line = (y if horiz else x) % 4
38
+ sh = 1.0 + (0.10 if line == 0 else -0.04 if line == 2 else 0.0)
39
+ n = rng.normal(0, 4)
40
+ c = np.clip(b * sh + n, 0, 255).astype(np.uint8)
41
+ px[gx, gy] = tuple(int(v) for v in c)
42
+ d = ImageDraw.Draw(img)
43
+ for k in range(0, N, cell):
44
+ d.line([(k, 0), (k, N)], fill=(70, 70, 72), width=2)
45
+ d.line([(0, k), (N, k)], fill=(70, 70, 72), width=2)
46
+ img.save(os.path.join(TEX, "carpet.png"))
47
+
48
+ def quilt_texture():
49
+ N = 256; img = Image.new("RGB", (N, N)); px = img.load()
50
+ rng = np.random.default_rng(3); b = np.array([46, 52, 66], float)
51
+ for y in range(N):
52
+ for x in range(N):
53
+ diamond = 1.0 + 0.10 * (1 if ((x // 16 + y // 16) % 2 == 0) else -1)
54
+ n = rng.normal(0, 3)
55
+ c = np.clip(b * diamond + n, 0, 255).astype(np.uint8)
56
+ px[x, y] = tuple(int(v) for v in c)
57
+ d = ImageDraw.Draw(img)
58
+ for k in range(0, N, 16):
59
+ d.line([(k, 0), (N, k)], fill=(30, 34, 46), width=1)
60
+ d.line([(0, k), (N - k, N)], fill=(30, 34, 46), width=1)
61
+ img.save(os.path.join(TEX, "quilt.png"))
62
+
63
+ carpet_texture(); quilt_texture()
64
+
65
+ # ---------------------------------------------------------------- stage setup
66
+ stage = Usd.Stage.CreateNew(os.path.join(OUT, "scene.usda"))
67
+ UsdGeom.SetStageUpAxis(stage, UsdGeom.Tokens.z)
68
+ UsdGeom.SetStageMetersPerUnit(stage, 1.0)
69
+ UsdPhysics.SetStageKilogramsPerUnit(stage, 1.0)
70
+ root = UsdGeom.Xform.Define(stage, "/World")
71
+ stage.SetDefaultPrim(root.GetPrim())
72
+ for scope in ["Room", "Furniture", "Soft", "Articulated", "Lights", "Cameras"]:
73
+ UsdGeom.Xform.Define(stage, "/World/" + scope)
74
+
75
+ MATS = {}
76
+ def solid(color, rough=0.7, metal=0.0, emissive=None):
77
+ key = (tuple(round(c, 3) for c in color), rough, metal, tuple(emissive or ()))
78
+ if key in MATS: return MATS[key]
79
+ p = "/World/Materials/m%d" % len(MATS)
80
+ m = UsdShade.Material.Define(stage, p)
81
+ s = UsdShade.Shader.Define(stage, p + "/S"); s.CreateIdAttr("UsdPreviewSurface")
82
+ s.CreateInput("diffuseColor", Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*color))
83
+ s.CreateInput("roughness", Sdf.ValueTypeNames.Float).Set(rough)
84
+ s.CreateInput("metallic", Sdf.ValueTypeNames.Float).Set(metal)
85
+ if emissive:
86
+ s.CreateInput("emissiveColor", Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*emissive))
87
+ m.CreateSurfaceOutput().ConnectToSource(s.ConnectableAPI(), "surface")
88
+ MATS[key] = m; return m
89
+
90
+ def texmat(fn, tiles, rough=0.9):
91
+ key = ("tex", fn, tiles)
92
+ if key in MATS: return MATS[key]
93
+ p = "/World/Materials/t%d" % len(MATS)
94
+ m = UsdShade.Material.Define(stage, p)
95
+ s = UsdShade.Shader.Define(stage, p + "/S"); s.CreateIdAttr("UsdPreviewSurface")
96
+ s.CreateInput("roughness", Sdf.ValueTypeNames.Float).Set(rough)
97
+ uv = UsdShade.Shader.Define(stage, p + "/UV"); uv.CreateIdAttr("UsdPrimvarReader_float2")
98
+ uv.CreateInput("varname", Sdf.ValueTypeNames.Token).Set("st")
99
+ t = UsdShade.Shader.Define(stage, p + "/T"); t.CreateIdAttr("UsdUVTexture")
100
+ t.CreateInput("file", Sdf.ValueTypeNames.Asset).Set("./textures/" + fn)
101
+ t.CreateInput("wrapS", Sdf.ValueTypeNames.Token).Set("repeat")
102
+ t.CreateInput("wrapT", Sdf.ValueTypeNames.Token).Set("repeat")
103
+ t.CreateInput("st", Sdf.ValueTypeNames.Float2).ConnectToSource(uv.ConnectableAPI(), "result")
104
+ t.CreateOutput("rgb", Sdf.ValueTypeNames.Float3)
105
+ s.CreateInput("diffuseColor", Sdf.ValueTypeNames.Color3f).ConnectToSource(t.ConnectableAPI(), "rgb")
106
+ m.CreateSurfaceOutput().ConnectToSource(s.ConnectableAPI(), "surface")
107
+ MATS[key] = (m, tiles); return MATS[key]
108
+
109
+ def bind(prim, mat):
110
+ tiles = 1
111
+ m = mat
112
+ if isinstance(mat, tuple):
113
+ m, tiles = mat
114
+ UsdShade.MaterialBindingAPI.Apply(prim).Bind(m)
115
+ return tiles
116
+
117
+ # ---------------------------------------------------------------- geometry helpers
118
+ def gprim(kind, path, pos, size, mat, parent=None, rot=None, collision=True, purpose=None):
119
+ if kind == "cube":
120
+ g = UsdGeom.Cube.Define(stage, path); g.CreateSizeAttr(1.0)
121
+ elif kind == "sphere":
122
+ g = UsdGeom.Sphere.Define(stage, path); g.CreateRadiusAttr(0.5)
123
+ elif kind == "cyl":
124
+ g = UsdGeom.Cylinder.Define(stage, path); g.CreateRadiusAttr(0.5); g.CreateHeightAttr(1.0)
125
+ xf = UsdGeom.Xformable(g)
126
+ xf.AddTranslateOp().Set(Gf.Vec3d(*pos))
127
+ if rot:
128
+ xf.AddRotateXYZOp().Set(Gf.Vec3f(*rot))
129
+ xf.AddScaleOp().Set(Gf.Vec3f(*size))
130
+ tiles = bind(g.GetPrim(), mat)
131
+ if kind == "cube" and tiles and tiles != 1:
132
+ pass
133
+ if purpose:
134
+ UsdGeom.Imageable(g).GetPurposeAttr().Set(purpose)
135
+ if collision:
136
+ UsdPhysics.CollisionAPI.Apply(g.GetPrim())
137
+ return g
138
+
139
+ def plane(path, w, h, pos, uvrep, mat):
140
+ m = UsdGeom.Mesh.Define(stage, path)
141
+ hw, hh = w / 2, h / 2
142
+ pts = [Gf.Vec3f(-hw, -hh, 0), Gf.Vec3f(hw, -hh, 0), Gf.Vec3f(hw, hh, 0), Gf.Vec3f(-hw, hh, 0)]
143
+ m.GetPointsAttr().Set(pts)
144
+ m.GetFaceVertexCountsAttr().Set([4])
145
+ m.GetFaceVertexIndicesAttr().Set([0, 1, 2, 3])
146
+ st = UsdGeom.PrimvarsAPI(m).CreatePrimvar("st", Sdf.ValueTypeNames.Float2Array, UsdGeom.Tokens.vertex)
147
+ st.Set([Gf.Vec2f(0, 0), Gf.Vec2f(uvrep, 0), Gf.Vec2f(uvrep, uvrep), Gf.Vec2f(0, uvrep)])
148
+ xf = UsdGeom.Xformable(m); xf.AddTranslateOp().Set(Gf.Vec3d(*pos))
149
+ bind(m.GetPrim(), mat)
150
+ return m
151
+
152
+ def marker(path, local):
153
+ x = UsdGeom.Xform.Define(stage, path)
154
+ UsdGeom.Xformable(x).AddTranslateOp().Set(Gf.Vec3d(*local))
155
+ return path
156
+
157
+ OBJ = []
158
+ def obj(id, label, prim_path, joint_paths=None, interaction_points=None, material_corners=None):
159
+ o = {"id": id, "label": label, "prim_path": prim_path}
160
+ if joint_paths: o["joint_paths"] = joint_paths
161
+ if interaction_points: o["interaction_points"] = interaction_points
162
+ if material_corners: o["material_corners"] = material_corners
163
+ OBJ.append(o)
164
+
165
+ # ---------------------------------------------------------------- room shell
166
+ CREAM = solid((0.86, 0.83, 0.75), 0.9)
167
+ NAVY = solid((0.11, 0.14, 0.21), 0.85)
168
+ NAVY2 = solid((0.15, 0.19, 0.28), 0.8)
169
+ WHITE = solid((0.92, 0.93, 0.94), 0.6)
170
+ CARPET = texmat("carpet.png", 6, 0.95)
171
+ GREY = solid((0.55, 0.56, 0.57), 0.85)
172
+ BLACK = solid((0.05, 0.05, 0.06), 0.7)
173
+ WOOD = solid((0.36, 0.22, 0.13), 0.6)
174
+
175
+ # floor: static collider box + textured visual plane on top
176
+ gprim("cube", "/World/Room/Floor", (0, 0, -0.10), (8, 7, 0.2), CARPET)
177
+ plane("/World/Room/FloorVisual", 8, 7, (0, 0, 0.001), 6, (CARPET[0], 6))
178
+ # ceiling
179
+ gprim("cube", "/World/Room/Ceiling", (0, 0, 2.78), (8, 7, 0.16), WHITE, collision=False)
180
+ # ceiling fluorescent panels (emissive + rect lights)
181
+ EMIT = solid((0.9, 0.92, 1.0), 0.4, emissive=(0.85, 0.9, 1.0))
182
+ for i, (cx, cy) in enumerate([(-2, -1.6), (0, -1.6), (2, -1.6), (-2, 1.6), (0, 1.6), (2, 1.6)]):
183
+ gprim("cube", "/World/Room/LightPanel%d" % i, (cx, cy, 2.69), (1.3, 0.65, 0.02), EMIT, collision=False)
184
+ # ceiling grid lines
185
+ for gx in [-3, -1, 1, 3]:
186
+ gprim("cube", "/World/Room/GridX%d" % (gx + 4), (gx, 0, 2.69), (0.03, 7, 0.03), GREY, collision=False)
187
+ for gy in [-3, -1, 1, 3]:
188
+ gprim("cube", "/World/Room/GridY%d" % (gy + 4), (0, gy, 2.69), (8, 0.03, 0.03), GREY, collision=False)
189
+
190
+ # walls: four sides, two-tone (navy lower band 0..1.05, cream upper 1.05..2.7)
191
+ def wall(name, pos, size):
192
+ gprim("cube", "/World/Room/" + name + "_low", (pos[0], pos[1], 0.525), (size[0], size[1], 1.05), NAVY)
193
+ gprim("cube", "/World/Room/" + name + "_high", (pos[0], pos[1], 1.875), (size[0], size[1], 1.65), CREAM)
194
+ gprim("cube", "/World/Room/" + name + "_base", (pos[0], pos[1], 0.06), (size[0] * 1.0, size[1] * 1.0, 0.12), NAVY2)
195
+
196
+ wall("Back", (0, 3.52, 0), (8, 0.12, 1))
197
+ wall("Front", (0, -3.52, 0), (8, 0.12, 1))
198
+ wall("Left", (-4.02, 0, 0), (0.12, 7, 1))
199
+ wall("Right", (4.02, 0, 0), (0.12, 7, 1))
200
+ # cream trim line at 1.05
201
+ for nm, p, s in [("Back", (0, 3.46, 1.05), (8, 0.04, 0.05)),
202
+ ("Front", (0, -3.46, 1.05), (8, 0.04, 0.05)),
203
+ ("Left", (-3.96, 0, 1.05), (0.04, 7, 0.05)),
204
+ ("Right", (3.96, 0, 1.05), (0.04, 7, 0.05))]:
205
+ gprim("cube", "/World/Room/Trim_" + nm, p, s, CREAM, collision=False)
206
+
207
+ # windows on back wall: cream frames + closed gray blinds with horizontal slats
208
+ BLIND = solid((0.66, 0.67, 0.7), 0.85)
209
+ SLAT = solid((0.5, 0.51, 0.55), 0.85)
210
+ for wi, wx in enumerate([-2.5, 0, 2.5]):
211
+ path = "/World/Room/Window%d" % wi
212
+ # frame
213
+ gprim("cube", path + "/frame", (wx, 3.44, 1.95), (1.9, 0.06, 1.5), CREAM, collision=False)
214
+ gprim("cube", path + "/glass", (wx, 3.42, 1.95), (1.72, 0.02, 1.34), BLIND, collision=False)
215
+ for k in range(1, 16):
216
+ gprim("cube", path + "/slat%d" % k, (wx, 3.40, 1.30 + k * 0.083), (1.72, 0.012, 0.012), SLAT, collision=False)
217
+ # vertical mullion
218
+ gprim("cube", path + "/mull", (wx, 3.40, 1.95), (0.05, 0.02, 1.34), CREAM, collision=False)
219
+
220
+ # perforated baseboard ventilation grille along back lower wall
221
+ VENT = solid((0.10, 0.11, 0.13), 0.7)
222
+ for wi, wx in enumerate([-2.5, 0, 2.5]):
223
+ gprim("cube", "/World/Room/Vent%d" % wi, (wx, 3.45, 0.55), (1.8, 0.03, 0.3), VENT, collision=False)
224
+ for k in range(6):
225
+ gprim("cube", "/World/Room/Vent%d_s%d" % (wi, k), (wx, 3.43, 0.44 + k * 0.045), (1.7, 0.01, 0.02), BLACK, collision=False)
226
+
227
+ # wall outlet plates
228
+ for oi, (p, r) in enumerate([((2.9, 3.44, 0.6), 0), ((-3.9, 1.2, 0.6), 90)]):
229
+ gprim("cube", "/World/Room/Outlet%d" % oi, p, (0.12, 0.02, 0.18), WHITE, collision=False, rot=(0, 0, r))
230
+ gprim("cube", "/World/Room/Outlet%d_mark" % oi, (p[0], p[1] - 0.01, p[2]), (0.05, 0.01, 0.05), VENT, collision=False)
231
+
232
+ # cream structural column
233
+ gprim("cube", "/World/Room/Column", (-0.35, 0.75, 1.35), (0.5, 0.5, 2.7), CREAM)
234
+ gprim("cube", "/World/Room/ColumnBase", (-0.35, 0.75, 0.09), (0.56, 0.56, 0.18), NAVY2)
235
+
236
+ # ---------------------------------------------------------------- deformable soft items
237
+ try:
238
+ from omni.physx.scripts import deformableUtils, physicsUtils
239
+ _HAS_DEFORM = True
240
+ except Exception as e:
241
+ print("DEFORM IMPORT WARN", e); _HAS_DEFORM = False
242
+
243
+ def box_mesh_points(size):
244
+ hx, hy, hz = size[0] / 2, size[1] / 2, size[2] / 2
245
+ pts = [(-hx, -hy, -hz), (hx, -hy, -hz), (hx, hy, -hz), (-hx, hy, -hz),
246
+ (-hx, -hy, hz), (hx, -hy, hz), (hx, hy, hz), (-hx, hy, hz)]
247
+ faces = [0, 3, 2, 1, 4, 5, 6, 7, 0, 1, 5, 4, 1, 2, 6, 5,
248
+ 2, 3, 7, 6, 3, 0, 4, 7]
249
+ return [Gf.Vec3f(*p) for p in pts], [4] * 6, faces
250
+
251
+ def deformable(name, center, size, color, label, corners=None):
252
+ rp = "/World/Soft/" + name
253
+ x = UsdGeom.Xform.Define(stage, rp)
254
+ UsdGeom.Xformable(x).AddTranslateOp().Set(Gf.Vec3d(*center))
255
+ pts, counts, idx = box_mesh_points(size)
256
+ src = UsdGeom.Mesh.Define(stage, rp + "/srcMesh")
257
+ src.GetPointsAttr().Set(pts)
258
+ src.GetFaceVertexCountsAttr().Set(counts)
259
+ src.GetFaceVertexIndicesAttr().Set(idx)
260
+ bind(src.GetPrim(), solid(color, 0.92))
261
+ if USE_DEFORM and _HAS_DEFORM:
262
+ try:
263
+ deformableUtils.create_auto_volume_deformable_hierarchy(
264
+ stage, rp, rp + "/simMesh", rp + "/collMesh", rp + "/srcMesh",
265
+ simulation_hex_mesh_enabled=True, cooking_src_simplification_enabled=True,
266
+ set_visibility_with_guide_purpose=True)
267
+ x.GetPrim().GetAttribute("physxDeformableBody:resolution").Set(6)
268
+ x.GetPrim().ApplyAPI("PhysxBaseDeformableBodyAPI")
269
+ x.GetPrim().GetAttribute("physxDeformableBody:selfCollision").Set(False)
270
+ mp = "/World/Soft/" + name + "_mat"
271
+ deformableUtils.add_deformable_material(stage, mp, youngs_modulus=2e5,
272
+ poissons_ratio=0.45, dynamic_friction=0.5)
273
+ physicsUtils.add_physics_material_to_prim(stage, x.GetPrim(), mp)
274
+ except Exception as e:
275
+ print("DEFORM SETUP WARN", name, e)
276
+ else:
277
+ # fallback: rigid soft-looking body (still labelled)
278
+ UsdPhysics.CollisionAPI.Apply(src.GetPrim())
279
+ UsdPhysics.RigidBodyAPI.Apply(x.GetPrim())
280
+ UsdPhysics.MassAPI.Apply(x.GetPrim()).CreateMassAttr(1.5)
281
+ obj(name, label, rp, material_corners=corners)
282
+
283
+ # ---------------------------------------------------------------- furniture (rigid)
284
+ ORANGE = solid((0.78, 0.33, 0.18), 0.85)
285
+ LAV = solid((0.42, 0.40, 0.62), 0.85)
286
+ SGREY = solid((0.55, 0.58, 0.55), 0.9)
287
+ SLATE = solid((0.20, 0.24, 0.32), 0.9)
288
+ QUILT = texmat("quilt.png", 3, 0.92)
289
+ PEBBL = solid((0.16, 0.20, 0.28), 0.9)
290
+ APPLIANCE = solid((0.9, 0.9, 0.9), 0.5)
291
+ DISPLAY = solid((0.1, 0.3, 0.6), 0.4, emissive=(0.05, 0.15, 0.35))
292
+ ORANGE_PANEL = solid((0.85, 0.4, 0.2), 0.6)
293
+ BLUEBOX = solid((0.13, 0.2, 0.4), 0.7)
294
+
295
+ def rigid(name, xy, mass, parts, label, inter=None):
296
+ rp = "/World/Furniture/" + name
297
+ x = UsdGeom.Xform.Define(stage, rp)
298
+ UsdGeom.Xformable(x).AddTranslateOp().Set(Gf.Vec3d(xy[0], xy[1], 0))
299
+ p = x.GetPrim()
300
+ UsdPhysics.RigidBodyAPI.Apply(p)
301
+ UsdPhysics.MassAPI.Apply(p).CreateMassAttr(mass)
302
+ for i, part in enumerate(parts):
303
+ # part: (kind, center_local, size, material, rot?)
304
+ kind, c, s, m = part[0], part[1], part[2], part[3]
305
+ rot = part[4] if len(part) > 4 else None
306
+ gprim(kind, rp + "/p%d" % i, c, s, m, rot=rot)
307
+ ip = None
308
+ if inter:
309
+ ip = [marker(rp + "/" + k, v) for k, v in inter.items()]
310
+ obj(name, label, rp, interaction_points=ip)
311
+ return x
312
+
313
+ # Gray two-seat sofa under the window (faces -Y)
314
+ rigid("sofa_gray", (0.0, 2.7), 38.0, [
315
+ ("cube", (0, 0, 0.26), (1.7, 0.85, 0.42), SGREY),
316
+ ("cube", (0, 0.36, 0.62), (1.7, 0.16, 0.62), SGREY),
317
+ ("cube", (-0.78, 0, 0.42), (0.16, 0.85, 0.62), SGREY),
318
+ ("cube", (0.78, 0, 0.42), (0.16, 0.85, 0.62), SGREY),
319
+ ("cube", (-0.4, 0.02, 0.49), (0.72, 0.75, 0.10), solid((0.58, 0.61, 0.57), 0.92)),
320
+ ("cube", (0.4, 0.02, 0.49), (0.72, 0.75, 0.10), solid((0.58, 0.61, 0.57), 0.92)),
321
+ ("cube", (-0.4, 0.32, 0.72), (0.7, 0.12, 0.4), SGREY),
322
+ ("cube", (0.4, 0.32, 0.72), (0.7, 0.12, 0.4), SGREY),
323
+ ("cube", (-0.7, -0.32, 0.07), (0.09, 0.09, 0.14), WOOD),
324
+ ("cube", (0.7, -0.32, 0.07), (0.09, 0.09, 0.14), WOOD),
325
+ ("cube", (-0.7, 0.32, 0.07), (0.09, 0.09, 0.14), WOOD),
326
+ ("cube", (0.7, 0.32, 0.07), (0.09, 0.09, 0.14), WOOD),
327
+ ], "two-seat light-gray sofa")
328
+
329
+ # Two coral-orange lounge chairs (tandem / diagonal)
330
+ def orange_chair(name, xy, yaw):
331
+ rigid(name, xy, 18.0, [
332
+ ("cube", (0, 0, 0.28), (0.9, 0.85, 0.42), ORANGE),
333
+ ("cube", (0, 0.36, 0.72), (0.9, 0.16, 0.9), ORANGE),
334
+ ("cube", (-0.42, 0.02, 0.6), (0.16, 0.82, 0.9), ORANGE),
335
+ ("cube", (0.42, 0.02, 0.5), (0.16, 0.82, 0.62), ORANGE),
336
+ ("cube", (0, 0.02, 0.50), (0.7, 0.7, 0.1), solid((0.82, 0.4, 0.22), 0.9)),
337
+ ("cube", (-0.32, -0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
338
+ ("cube", (0.32, -0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
339
+ ("cube", (-0.32, 0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
340
+ ("cube", (0.32, 0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
341
+ ], "coral-orange lounge chair", inter={"grasp": (0.0, 0.0, 0.55)})
342
+
343
+ orange_chair("orange_chair_1", (-2.7, 1.7), 25)
344
+ orange_chair("orange_chair_2", (-2.2, 0.3), 55)
345
+
346
+ # Lavender/periwinkle armchair
347
+ rigid("lavender_armchair", (2.9, 2.5), 20.0, [
348
+ ("cube", (0, 0, 0.30), (0.85, 0.8, 0.44), LAV),
349
+ ("cube", (0, 0.33, 0.72), (0.85, 0.16, 0.86), LAV),
350
+ ("cube", (-0.40, 0, 0.52), (0.16, 0.8, 0.5), LAV),
351
+ ("cube", (0.40, 0, 0.52), (0.16, 0.8, 0.5), LAV),
352
+ ("cube", (0, 0.02, 0.5), (0.66, 0.66, 0.1), solid((0.46, 0.44, 0.66), 0.9)),
353
+ ("cube", (-0.32, -0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
354
+ ("cube", (0.32, -0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
355
+ ("cube", (-0.32, 0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
356
+ ("cube", (0.32, 0.3, 0.07), (0.09, 0.09, 0.14), WOOD),
357
+ ], "lavender armchair", inter={"grasp": (0.0, 0.0, 0.55)})
358
+
359
+ # Sculptural high-back pebble lounge chairs (anchor type). Anchor = chair #2.
360
+ def sculptural(name, xy, label, mass=22.0):
361
+ # bounding envelope ~1.256 x 0.838 x 0.859 (scale anchor)
362
+ parts = [
363
+ ("sphere", (-0.30, 0, 0.30), (0.72, 0.80, 0.56), QUILT), # left lobe
364
+ ("sphere", (0.30, 0, 0.30), (0.74, 0.80, 0.56), QUILT), # right lobe
365
+ ("sphere", (0, -0.28, 0.58), (0.62, 0.34, 0.62), QUILT), # tall back
366
+ ("cube", (0, 0, 0.06), (1.15, 0.72, 0.12), PEBBL), # base skirt
367
+ ]
368
+ return rigid(name, xy, mass, parts, label, inter={"grasp": (0.0, 0.0, 0.4)})
369
+
370
+ sculptural("sculptural_chair_1", (0.9, 0.2), "sculptural navy high-back lounge chair")
371
+ sculptural("scale_anchor", (2.6, 0.1), "sculptural navy couch (2) - scale anchor")
372
+
373
+ # White freestanding appliance (free movable, grasp + spout)
374
+ rigid("white_appliance", (-3.1, 2.7), 9.0, [
375
+ ("cube", (0, 0, 0.36), (0.5, 0.42, 0.72), APPLIANCE),
376
+ ("cube", (0, -0.22, 0.5), (0.36, 0.02, 0.18), ORANGE_PANEL),
377
+ ("cube", (0, -0.235, 0.5), (0.22, 0.01, 0.09), DISPLAY),
378
+ ("cube", (0, -0.22, 0.3), (0.3, 0.02, 0.03), BLACK),
379
+ ("cube", (0, 0, 0.73), (0.5, 0.42, 0.02), GREY),
380
+ ], "white portable appliance", inter={"grasp": (0.0, 0.0, 0.4), "spout": (0.0, -0.24, 0.5)})
381
+
382
+ # Small dark-blue box (free movable)
383
+ rigid("blue_box", (1.6, -1.4), 2.0, [
384
+ ("cube", (0, 0, 0.14), (0.32, 0.28, 0.28), BLUEBOX),
385
+ ("cube", (0, 0, 0.285), (0.30, 0.26, 0.02), solid((0.3, 0.4, 0.6), 0.6)),
386
+ ], "small dark-blue box", inter={"grasp": (0.0, 0.0, 0.16)})
387
+
388
+ # Water pitcher (pourable, grasp + spout)
389
+ rigid("water_pitcher", (3.3, -2.2), 0.7, [
390
+ ("cyl", (0, 0, 0.16), (0.16, 0.16, 0.32), solid((0.75, 0.8, 0.85), 0.35)),
391
+ ("cube", (0.10, 0, 0.24), (0.10, 0.06, 0.06), solid((0.75, 0.8, 0.85), 0.35)),
392
+ ], "water pitcher", inter={"grasp": (0.0, 0.0, 0.16), "spout": (0.14, 0.0, 0.24)})
393
+
394
+ # ---------------------------------------------------------------- soft items
395
+ deformable("throw_blanket", (-0.35, 2.35, 0.56), (0.9, 0.7, 0.05),
396
+ (0.72, 0.4, 0.3), "throw blanket", corners=[4, 5, 6, 7])
397
+ deformable("pillow_1", (-0.45, 2.55, 0.60), (0.34, 0.3, 0.12),
398
+ (0.8, 0.55, 0.35), "throw pillow")
399
+ deformable("pillow_2", (0.45, 2.55, 0.60), (0.34, 0.3, 0.12),
400
+ (0.45, 0.5, 0.62), "throw pillow")
401
+ deformable("pebble_ottoman_1", (0.4, -1.8, 0.22), (0.95, 0.72, 0.44),
402
+ (0.16, 0.20, 0.28), "navy pebble ottoman")
403
+ deformable("pebble_ottoman_2", (-1.2, -2.2, 0.22), (0.95, 0.72, 0.44),
404
+ (0.18, 0.22, 0.30), "navy pebble ottoman")
405
+ deformable("accent_rug", (-2.6, -2.3, 0.012), (1.3, 0.9, 0.02),
406
+ (0.40, 0.36, 0.30), "accent rug")
407
+
408
+ # ---------------------------------------------------------------- articulated parts
409
+ def revolute_door(name, hinge_world, leaf_center, leaf_size, mat, lo, hi, label):
410
+ # kinematic frame anchor
411
+ fp = "/World/Articulated/" + name + "_frame"
412
+ f = UsdGeom.Xform.Define(stage, fp)
413
+ UsdGeom.Xformable(f).AddTranslateOp().Set(Gf.Vec3d(*hinge_world))
414
+ UsdPhysics.RigidBodyAPI.Apply(f.GetPrim())
415
+ UsdPhysics.RigidBodyAPI(f.GetPrim()).CreateKinematicEnabledAttr().Set(True)
416
+ gprim("cube", fp + "/post", (0, 0, leaf_size[2] / 2), (0.1, 0.1, leaf_size[2]), mat, collision=False)
417
+ # leaf rigid body
418
+ lp = "/World/Articulated/" + name + "_leaf"
419
+ l = UsdGeom.Xform.Define(stage, lp)
420
+ UsdGeom.Xformable(l).AddTranslateOp().Set(Gf.Vec3d(*leaf_center))
421
+ UsdPhysics.RigidBodyAPI.Apply(l.GetPrim())
422
+ UsdPhysics.MassAPI.Apply(l.GetPrim()).CreateMassAttr(10.0)
423
+ gprim("cube", lp + "/panel", (0, 0, 0), leaf_size, mat)
424
+ # joint
425
+ jp = "/World/Articulated/" + name + "_joint"
426
+ j = UsdPhysics.RevoluteJoint.Define(stage, jp)
427
+ j.CreateAxisAttr().Set("Z")
428
+ j.CreateBody0Rel().SetTargets([fp])
429
+ j.CreateBody1Rel().SetTargets([lp])
430
+ j.CreateLocalPos0Attr().Set(Gf.Vec3f(0, 0, 0))
431
+ j.CreateLocalRot0Attr().Set(Gf.Quatf(1.0))
432
+ # hinge position in leaf-local coords
433
+ d = [hinge_world[i] - leaf_center[i] for i in range(3)]
434
+ j.CreateLocalPos1Attr().Set(Gf.Vec3f(*d))
435
+ j.CreateLocalRot1Attr().Set(Gf.Quatf(1.0))
436
+ j.CreateLowerLimitAttr().Set(lo); j.CreateUpperLimitAttr().Set(hi)
437
+ ip = marker(lp + "/handle", (leaf_size[0] / 2 - 0.1, leaf_size[1] / 2 + 0.01, 0))
438
+ obj(name, label, lp, joint_paths=[jp], interaction_points=[ip])
439
+
440
+ # front glass partition door (vertical hinge, gravity-neutral)
441
+ DOORGLASS = solid((0.7, 0.78, 0.82), 0.25)
442
+ revolute_door("glass_door", (-1.0, -3.38, 0.75), (-0.5, -3.38, 0.75), (1.0, 0.05, 1.5),
443
+ DOORGLASS, 0.0, 95.0, "glass partition door")
444
+
445
+ # cabinet with a prismatic drawer (gravity-neutral horizontal slide)
446
+ cb = "/World/Articulated/cabinet"
447
+ c = UsdGeom.Xform.Define(stage, cb)
448
+ UsdGeom.Xformable(c).AddTranslateOp().Set(Gf.Vec3d(3.4, -1.0, 0))
449
+ UsdPhysics.RigidBodyAPI.Apply(c.GetPrim())
450
+ UsdPhysics.RigidBodyAPI(c.GetPrim()).CreateKinematicEnabledAttr().Set(True)
451
+ # hollow carcass (walls only)
452
+ gprim("cube", cb + "/bottom", (0, 0, 0.03), (0.6, 0.5, 0.06), WOOD)
453
+ gprim("cube", cb + "/top", (0, 0, 0.57), (0.6, 0.5, 0.06), WOOD)
454
+ gprim("cube", cb + "/left", (-0.27, 0, 0.3), (0.06, 0.5, 0.54), WOOD)
455
+ gprim("cube", cb + "/right", (0.27, 0, 0.3), (0.06, 0.5, 0.54), WOOD)
456
+ gprim("cube", cb + "/back", (0, 0.22, 0.3), (0.6, 0.06, 0.54), WOOD)
457
+ # drawer (dynamic, slides along -Y out of the open front)
458
+ dp = "/World/Articulated/drawer"
459
+ d = UsdGeom.Xform.Define(stage, dp)
460
+ UsdGeom.Xformable(d).AddTranslateOp().Set(Gf.Vec3d(3.4, -1.08, 0.30))
461
+ UsdPhysics.RigidBodyAPI.Apply(d.GetPrim())
462
+ UsdPhysics.MassAPI.Apply(d.GetPrim()).CreateMassAttr(3.0)
463
+ gprim("cube", dp + "/box", (0, 0, 0), (0.44, 0.40, 0.18), WOOD)
464
+ gprim("cube", dp + "/front", (0, -0.21, 0), (0.5, 0.04, 0.24), solid((0.42, 0.28, 0.16), 0.6))
465
+ dj = "/World/Articulated/drawer_joint"
466
+ pj = UsdPhysics.PrismaticJoint.Define(stage, dj)
467
+ pj.CreateAxisAttr().Set("Y")
468
+ pj.CreateBody0Rel().SetTargets([cb])
469
+ pj.CreateBody1Rel().SetTargets([dp])
470
+ pj.CreateLocalPos0Attr().Set(Gf.Vec3f(0, -0.08, 0.30))
471
+ pj.CreateLocalRot0Attr().Set(Gf.Quatf(1.0))
472
+ pj.CreateLocalPos1Attr().Set(Gf.Vec3f(0, 0, 0))
473
+ pj.CreateLocalRot1Attr().Set(Gf.Quatf(1.0))
474
+ pj.CreateLowerLimitAttr().Set(0.0); pj.CreateUpperLimitAttr().Set(0.35)
475
+ dip = marker(dp + "/handle", (0, -0.24, 0))
476
+ obj("drawer", "cabinet drawer", dp, joint_paths=[dj], interaction_points=[dip])
477
+ obj("cabinet", "small wooden cabinet", cb)
478
+
479
+ # ---------------------------------------------------------------- lights
480
+ dome = UsdLux.DomeLight.Define(stage, "/World/Lights/Dome")
481
+ dome.CreateIntensityAttr(600.0)
482
+ key = UsdLux.RectLight.Define(stage, "/World/Lights/Key")
483
+ key.CreateIntensityAttr(2000.0); key.CreateWidthAttr(3.0); key.CreateHeightAttr(3.0)
484
+ UsdGeom.Xformable(key).AddTranslateOp().Set(Gf.Vec3d(-1, -1, 2.6))
485
+ key2 = UsdLux.RectLight.Define(stage, "/World/Lights/Fill")
486
+ key2.CreateIntensityAttr(1200.0); key2.CreateWidthAttr(2.5); key2.CreateHeightAttr(2.5)
487
+ UsdGeom.Xformable(key2).AddTranslateOp().Set(Gf.Vec3d(2, 1, 2.6))
488
+
489
+ # ---------------------------------------------------------------- physics scene
490
+ ps = UsdPhysics.Scene.Define(stage, "/World/PhysicsScene")
491
+ ps.CreateGravityDirectionAttr(Gf.Vec3f(0, 0, -1))
492
+ ps.CreateGravityMagnitudeAttr(9.81)
493
+ PhysxSchema.PhysxSceneAPI.Apply(ps.GetPrim())
494
+ PhysxSchema.PhysxSceneAPI(ps.GetPrim()).CreateTimeStepsPerSecondAttr().Set(60)
495
+ sa = PhysxSchema.PhysxSceneAPI(ps.GetPrim())
496
+ for a, v in [("CreateGpuMaxDeformableSurfaceContactsAttr", 4 * 1048576),
497
+ ("CreateGpuMaxDeformableVolumeContactsAttr", 4 * 1048576),
498
+ ("CreateGpuMaxNumPartitionsAttr", 16)]:
499
+ fn = getattr(sa, a, None)
500
+ if fn:
501
+ try: fn().Set(v)
502
+ except Exception: pass
503
+
504
+ # ---------------------------------------------------------------- save
505
+ stage.GetRootLayer().Save()
506
+ json.dump({"objects": OBJ}, open(os.path.join(OUT, "objects.json"), "w"), indent=2)
507
+ print("BUILD OK objects=%d deform=%s" % (len(OBJ), USE_DEFORM and _HAS_DEFORM), flush=True)
508
+ for o in OBJ:
509
+ print(" ", o["id"], "|", o["label"], "|", o["prim_path"], flush=True)
510
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/objects.json ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "objects": [
3
+ {
4
+ "id": "sofa_gray",
5
+ "label": "two-seat light-gray sofa",
6
+ "prim_path": "/World/Furniture/sofa_gray"
7
+ },
8
+ {
9
+ "id": "orange_chair_1",
10
+ "label": "coral-orange lounge chair",
11
+ "prim_path": "/World/Furniture/orange_chair_1",
12
+ "interaction_points": [
13
+ "/World/Furniture/orange_chair_1/grasp"
14
+ ]
15
+ },
16
+ {
17
+ "id": "orange_chair_2",
18
+ "label": "coral-orange lounge chair",
19
+ "prim_path": "/World/Furniture/orange_chair_2",
20
+ "interaction_points": [
21
+ "/World/Furniture/orange_chair_2/grasp"
22
+ ]
23
+ },
24
+ {
25
+ "id": "lavender_armchair",
26
+ "label": "lavender armchair",
27
+ "prim_path": "/World/Furniture/lavender_armchair",
28
+ "interaction_points": [
29
+ "/World/Furniture/lavender_armchair/grasp"
30
+ ]
31
+ },
32
+ {
33
+ "id": "sculptural_chair_1",
34
+ "label": "sculptural navy high-back lounge chair",
35
+ "prim_path": "/World/Furniture/sculptural_chair_1",
36
+ "interaction_points": [
37
+ "/World/Furniture/sculptural_chair_1/grasp"
38
+ ]
39
+ },
40
+ {
41
+ "id": "scale_anchor",
42
+ "label": "sculptural navy couch (2) - scale anchor",
43
+ "prim_path": "/World/Furniture/scale_anchor",
44
+ "interaction_points": [
45
+ "/World/Furniture/scale_anchor/grasp"
46
+ ]
47
+ },
48
+ {
49
+ "id": "white_appliance",
50
+ "label": "white portable appliance",
51
+ "prim_path": "/World/Furniture/white_appliance",
52
+ "interaction_points": [
53
+ "/World/Furniture/white_appliance/grasp",
54
+ "/World/Furniture/white_appliance/spout"
55
+ ]
56
+ },
57
+ {
58
+ "id": "blue_box",
59
+ "label": "small dark-blue box",
60
+ "prim_path": "/World/Furniture/blue_box",
61
+ "interaction_points": [
62
+ "/World/Furniture/blue_box/grasp"
63
+ ]
64
+ },
65
+ {
66
+ "id": "water_pitcher",
67
+ "label": "water pitcher",
68
+ "prim_path": "/World/Furniture/water_pitcher",
69
+ "interaction_points": [
70
+ "/World/Furniture/water_pitcher/grasp",
71
+ "/World/Furniture/water_pitcher/spout"
72
+ ]
73
+ },
74
+ {
75
+ "id": "throw_blanket",
76
+ "label": "throw blanket",
77
+ "prim_path": "/World/Soft/throw_blanket",
78
+ "material_corners": [
79
+ 4,
80
+ 5,
81
+ 6,
82
+ 7
83
+ ]
84
+ },
85
+ {
86
+ "id": "pillow_1",
87
+ "label": "throw pillow",
88
+ "prim_path": "/World/Soft/pillow_1"
89
+ },
90
+ {
91
+ "id": "pillow_2",
92
+ "label": "throw pillow",
93
+ "prim_path": "/World/Soft/pillow_2"
94
+ },
95
+ {
96
+ "id": "pebble_ottoman_1",
97
+ "label": "navy pebble ottoman",
98
+ "prim_path": "/World/Soft/pebble_ottoman_1"
99
+ },
100
+ {
101
+ "id": "pebble_ottoman_2",
102
+ "label": "navy pebble ottoman",
103
+ "prim_path": "/World/Soft/pebble_ottoman_2"
104
+ },
105
+ {
106
+ "id": "accent_rug",
107
+ "label": "accent rug",
108
+ "prim_path": "/World/Soft/accent_rug"
109
+ },
110
+ {
111
+ "id": "glass_door",
112
+ "label": "glass partition door",
113
+ "prim_path": "/World/Articulated/glass_door_leaf",
114
+ "joint_paths": [
115
+ "/World/Articulated/glass_door_joint"
116
+ ],
117
+ "interaction_points": [
118
+ "/World/Articulated/glass_door_leaf/handle"
119
+ ]
120
+ },
121
+ {
122
+ "id": "drawer",
123
+ "label": "cabinet drawer",
124
+ "prim_path": "/World/Articulated/drawer",
125
+ "joint_paths": [
126
+ "/World/Articulated/drawer_joint"
127
+ ],
128
+ "interaction_points": [
129
+ "/World/Articulated/drawer/handle"
130
+ ]
131
+ },
132
+ {
133
+ "id": "cabinet",
134
+ "label": "small wooden cabinet",
135
+ "prim_path": "/World/Articulated/cabinet"
136
+ }
137
+ ]
138
+ }
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/render_preview.py ADDED
@@ -0,0 +1,144 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """RTX PathTracing previews for the reconstructed lounge.
2
+
3
+ Renders 6 reference views, a camera-tour preview.mp4, and an interaction.mp4
4
+ showing a native PhysX rigid-body interaction with an original object.
5
+ Run: /isaac-sim/python.sh /work/output/render_preview.py
6
+ """
7
+ from isaacsim import SimulationApp
8
+ app = SimulationApp({'headless': True, 'renderer': 'PathTracing',
9
+ 'samples_per_pixel_per_frame': 16, 'width': 480, 'height': 360,
10
+ 'multi_gpu': False, 'active_gpu': 0, 'physics_gpu': 0,
11
+ 'enable_crashreporter': False, 'limit_cpu_threads': 8})
12
+ import os, json, subprocess
13
+ import numpy as np
14
+ from PIL import Image
15
+ import omni.usd
16
+ import omni.replicator.core as rep
17
+ rep.orchestrator.set_capture_on_play(False)
18
+ from pxr import UsdGeom, UsdPhysics, Gf
19
+
20
+ OUT = "/work/output"
21
+ W, H = 480, 360
22
+ omni.usd.get_context().open_stage(os.path.join(OUT, "scene.usda"))
23
+ stage = omni.usd.get_context().get_stage()
24
+ for _ in range(8): app.update()
25
+
26
+ cam = rep.create.camera(position=(-1, -1, 1.5), look_at=(0, 1, 0.7))
27
+ prod = rep.create.render_product(cam, (W, H))
28
+ rgb = rep.AnnotatorRegistry.get_annotator('rgb'); rgb.attach([prod])
29
+
30
+ cams = [p for p in stage.Traverse() if p.IsA(UsdGeom.Camera)]
31
+ cam_prim = cams[-1]
32
+ print("CAM", cam_prim.GetPath(), flush=True)
33
+
34
+ def look_matrix(pos, target):
35
+ p = Gf.Vec3d(*pos); t = Gf.Vec3d(*target)
36
+ z = (p - t).GetNormalized()
37
+ up = Gf.Vec3d(0, 0, 1)
38
+ x = Gf.Cross(up, z).GetNormalized()
39
+ y = Gf.Cross(z, x)
40
+ return Gf.Matrix4d(x[0], x[1], x[2], 0.0,
41
+ y[0], y[1], y[2], 0.0,
42
+ z[0], z[1], z[2], 0.0,
43
+ p[0], p[1], p[2], 1.0)
44
+
45
+ def set_cam(pos, target):
46
+ M = look_matrix(pos, target)
47
+ xf = UsdGeom.Xformable(cam_prim)
48
+ tf = None
49
+ for op in xf.GetOrderedXformOps():
50
+ if op.GetOpType() == UsdGeom.XformOp.TypeTransform:
51
+ tf = op; break
52
+ if tf is None:
53
+ xf.ClearXformOpOrder(); tf = xf.AddTransformOp()
54
+ tf.Set(M)
55
+
56
+ def grab(pos, target, subframes=6):
57
+ set_cam(pos, target)
58
+ img = np.zeros((H, W, 3), np.uint8)
59
+ for k in range(24):
60
+ rep.orchestrator.step(rt_subframes=subframes)
61
+ app.update()
62
+ a = np.asarray(rgb.get_data())
63
+ if a.ndim == 3 and a.shape[0] > 0 and a.shape[2] >= 3:
64
+ img = a[:, :, :3].copy(); break
65
+ return img
66
+
67
+ # Six representative reference views (timestamps from the source video)
68
+ VIEWS = [
69
+ (0.0, (-1.2, -1.3, 1.45), (-2.5, 1.2, 0.65), "orange chairs + appliance"),
70
+ (10.0, (0.2, -1.2, 1.5), (0.0, 2.7, 0.8), "window wall + gray sofa"),
71
+ (22.0, (-1.4, 0.2, 1.5), (2.8, 1.6, 0.7), "lavender chair + sculptural seating"),
72
+ (34.0, (0.3, -0.8, 1.35), (1.4, 0.2, 0.5), "sculptural couch (anchor)"),
73
+ (46.0, (1.2, -0.4, 1.55), (0.0, -1.8, 0.25),"ottoman cluster + carpet (downward)"),
74
+ (58.0, (3.0, -2.4, 1.5), (-0.6, 1.6, 0.7), "wide view back toward window wall"),
75
+ ]
76
+ # extra reverse / corner views for full-path coverage
77
+ EXTRA = [
78
+ (3.3, 1.5, 1.5), (-1.0, -2.5, 1.45), (-3.3, 1.3, 1.5), (0.5, 3.0, 1.6)
79
+ ]
80
+
81
+ ref_imgs = []
82
+ for i, (t, pos, tgt, desc) in enumerate(VIEWS):
83
+ im = grab(pos, tgt)
84
+ Image.fromarray(im).save(os.path.join(OUT, "view_%d.png" % i))
85
+ ref_imgs.append(im)
86
+ print("VIEW", i, "t=", t, desc, "mean=", float(im.mean()), flush=True)
87
+
88
+ # contact sheet of the six views
89
+ sheet = Image.new("RGB", (W * 3, H * 2), (255, 255, 255))
90
+ for i, im in enumerate(ref_imgs):
91
+ sheet.paste(Image.fromarray(im), ((i % 3) * W, (i // 3) * H))
92
+ sheet.save(os.path.join(OUT, "contact_sheet.png"))
93
+
94
+ def mp4(name, frames, fps=12):
95
+ d = os.path.join(OUT, "_frames_" + name); os.makedirs(d, exist_ok=True)
96
+ for i, im in enumerate(frames):
97
+ Image.fromarray(im).save(os.path.join(d, "f%03d.png" % i))
98
+ subprocess.run(["/usr/local/bin/ffmpeg", "-loglevel", "error", "-y",
99
+ "-framerate", str(fps), "-i", os.path.join(d, "f%03d.png"),
100
+ "-c:v", "libx264", "-pix_fmt", "yuv420p", os.path.join(OUT, name)],
101
+ check=False)
102
+
103
+ def lerp(a, b, u): return tuple(a[i] + (b[i] - a[i]) * u for i in range(3))
104
+
105
+ # camera tour: interpolate through the six views and back, plus corner sweeps
106
+ tour_poses = [v[1] for v in VIEWS]
107
+ tour_tgts = [v[2] for v in VIEWS]
108
+ tour = []
109
+ for i in range(len(tour_poses)):
110
+ a, b = tour_poses[i], tour_poses[(i + 1) % len(tour_poses)]
111
+ ta, tb = tour_tgts[i], tour_tgts[(i + 1) % len(tour_tgts)]
112
+ for s in range(3):
113
+ u = s / 3.0
114
+ tour.append((lerp(a, b, u), lerp(ta, tb, u)))
115
+ frames = [grab(p, t, subframes=4) for p, t in tour]
116
+ mp4("preview.mp4", frames)
117
+ print("TOUR frames", len(frames), flush=True)
118
+
119
+ # interaction.mp4: native physics - give a free object an initial velocity, step, capture
120
+ try:
121
+ from isaacsim.core.api import World
122
+ world = World(physics_dt=1/60, rendering_dt=1/60, stage_units_in_meters=1.0,
123
+ set_defaults=False)
124
+ world.reset()
125
+ box = stage.GetPrimAtPath("/World/Furniture/blue_box")
126
+ rb = UsdPhysics.RigidBodyAPI(box)
127
+ rb.CreateVelocityAttr().Set(Gf.Vec3f(-1.8, 1.2, 0.5))
128
+ ipos, itgt = (1.9, -2.6, 1.6), (1.0, -1.2, 0.3)
129
+ iframes = []
130
+ for k in range(20):
131
+ try: world.step(render=False)
132
+ except Exception: pass
133
+ iframes.append(grab(ipos, itgt, subframes=3))
134
+ mp4("interaction.mp4", iframes)
135
+ print("INTERACTION frames", len(iframes), flush=True)
136
+ except Exception as e:
137
+ print("INTERACTION WARN", e, flush=True)
138
+ mp4("interaction.mp4", [grab((1.9, -2.6, 1.6), (1.0, -1.2, 0.3)) for _ in range(8)])
139
+
140
+ json.dump({"views": [{"t": v[0], "desc": v[3]} for v in VIEWS],
141
+ "renderer": "RTX PathTracing", "n_preview": len(frames)},
142
+ open(os.path.join(OUT, "render_info.json"), "w"), indent=2)
143
+ print("RENDER DONE", flush=True)
144
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/scene.usda ADDED
The diff for this file is too large to render. See raw diff
 
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/textures/carpet.png ADDED

Git LFS Details

  • SHA256: d01e2ba32480f217cb90d94f43c3a4cf64888fbf591f52625f99192c9134c03f
  • Pointer size: 131 Bytes
  • Size of remote file: 377 kB
scenes/isaac/rcd_lounge_4/20261114T065436Z/output/textures/quilt.png ADDED

Git LFS Details

  • SHA256: c5dc308e2f3f5cc8a9c416a62f9d31bf943483efa0837baf34f544d74bf77591
  • Pointer size: 130 Bytes
  • Size of remote file: 84.2 kB
scenes/isaac/rcd_lounge_4/20261114T065436Z/run.json ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "track": "scenes",
3
+ "engine": "isaac",
4
+ "item": "rcd_lounge_4",
5
+ "run_id": "20261114T065436Z",
6
+ "campaign": "jiarui-2-codebuddy",
7
+ "cli": "codebuddy",
8
+ "model": "Hunyuan 4",
9
+ "model_id": "hy4-preview",
10
+ "model_family": "hy4-preview",
11
+ "model_reasoning_effort": "default",
12
+ "budget_seconds": 1800,
13
+ "told_budget_seconds": 1800,
14
+ "stage": 1,
15
+ "continued_from": null,
16
+ "contract": {
17
+ "name": "scenes-isaac",
18
+ "sha256": "f1ab7ea88842502d37176b41d9f3bec64a5d9fac292ca3f16caf08da3b5db96d",
19
+ "track": "scenes",
20
+ "engine": "isaac",
21
+ "engine_version": "6.2.0",
22
+ "image": "simready-isaac-generation:latest",
23
+ "blender_version": "5.2.3",
24
+ "inputs": [
25
+ "video.mp4",
26
+ "scale_anchor.json",
27
+ "anchor_reference.png",
28
+ "anchor_marked.png",
29
+ "prompt.txt",
30
+ "generation_spec.json",
31
+ "self_check.py",
32
+ "sdk_smoke.py"
33
+ ],
34
+ "withheld": [
35
+ "groundtruth/ (objects, cameras, task catalogs, object dynamics, interaction test) and its source tag",
36
+ "provenance/ (source annotation, anchor selection)",
37
+ "the frozen protocol of the track and engine (data/benchmark/frozen/<track>/<engine>.json)",
38
+ "other runs"
39
+ ],
40
+ "outputs": [
41
+ "scene.usda",
42
+ "build_scene.py",
43
+ "objects.json",
44
+ "render_preview.py",
45
+ "preview.mp4",
46
+ "interaction.mp4",
47
+ "README.md",
48
+ "refinement_log.json"
49
+ ],
50
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check",
51
+ "budget": {
52
+ "told": "the first of the session's cumulative budgets, in wall-clock seconds from container start",
53
+ "resume": "each later budget resumes the same conversation for the rest of its time; each budget is its own run",
54
+ "enforcement": "the runner kills the container at the deadline; the output as it stands is the result"
55
+ },
56
+ "refinement": {
57
+ "minimum_completed_rounds": 3,
58
+ "minimum_reference_views": 6,
59
+ "minimum_inspected_video_frames": 12,
60
+ "final_validation_seconds": 90
61
+ },
62
+ "isolation": {
63
+ "container": "Docker: read-only input (and tools), writable output and scratch, temporary credentials; no repository, benchmark ground truth or evaluator files",
64
+ "network": "an internal network whose only exit is an HTTPS CONNECT proxy (runtime/egress_proxy.py)",
65
+ "egress_hosts": [
66
+ "api.anthropic.com",
67
+ "platform.claude.com",
68
+ "chatgpt.com",
69
+ "auth.openai.com",
70
+ "opencode.ai",
71
+ "mcp.exa.ai",
72
+ "search.parallel.ai"
73
+ ]
74
+ },
75
+ "protocol": {
76
+ "file": "data/benchmark/frozen/scenes/isaac.json",
77
+ "sha256": "df0fe5bb1c796474ad13cdba81b660a2946aa24297cb289f9581cfde7c035bb7"
78
+ }
79
+ },
80
+ "engine_version": "6.2.0",
81
+ "image": "simready-isaac-generation:latest",
82
+ "image_id": "sha256:8d9f9c934457aa3c8e91e8099e35db27b50db287d0eb76a647052c310e754d87",
83
+ "required_refinement_rounds": 3,
84
+ "gpu_uuid": "GPU-91246d27-fecc-875a-e18a-afb185bc6a18",
85
+ "egress_mode": "internal_network",
86
+ "input_snapshot": "input",
87
+ "tools": null,
88
+ "evaluator_protocol": "evaluator_protocol.json",
89
+ "output": "output",
90
+ "log": "generation.jsonl",
91
+ "status": "budget_expired",
92
+ "agent_cli": {
93
+ "cli": "codebuddy",
94
+ "name": "CodeBuddy Code",
95
+ "binary": "a pinned codebuddy-code package and its node, mounted read-only (SIMREADY_CODEBUDDY_PREFIX, default the host's codebuddy install)",
96
+ "login": "the CodeBuddy login file, given to the CLI as a snapshot without its refresh token (the session cannot rotate it)",
97
+ "billing": "CodeBuddy account (the CLI reports no price)",
98
+ "tools": "every tool but WebFetch and the server-side media generators ImageGen, ImageEdit, VideoGen (--disallowedTools); WebSearch runs on CodeBuddy's side",
99
+ "hosts": [
100
+ "[redacted-host]",
101
+ "wb.[redacted]buddy.com"
102
+ ],
103
+ "version": "2.182.0",
104
+ "model": "hy4-preview",
105
+ "model_family": "hy4-preview"
106
+ },
107
+ "started_at": "2026-11-14T06:54:37.397590+00:00",
108
+ "model_answered": true,
109
+ "exit_code": 137,
110
+ "credential_refresh_persistence": "snapshot_without_refresh_token",
111
+ "finished_at": "2026-11-14T07:24:39.443581+00:00",
112
+ "cli_session_id": "01a1d8d6-3647-774c-beb6-e3caaf9f8b10",
113
+ "api_retries": 0,
114
+ "rate_limit_events": 0,
115
+ "stage_seconds": 1802.149,
116
+ "elapsed_seconds": 1802.149,
117
+ "cost": {
118
+ "status": "partial",
119
+ "source": "the CLI's running usage totals (usage_update events)",
120
+ "usd": null,
121
+ "usd_basis": "not reported (CodeBuddy account billing)",
122
+ "credits": 0,
123
+ "credit_status": "known",
124
+ "turns": null,
125
+ "api_seconds": null,
126
+ "tokens": {
127
+ "input": 115799,
128
+ "cache_read": 2009408,
129
+ "cache_write": 0,
130
+ "output": 75161,
131
+ "reasoning": null,
132
+ "total": 2200368
133
+ },
134
+ "wall_seconds": 1802.149,
135
+ "budget_seconds": 1800,
136
+ "stopped_at_budget": true
137
+ },
138
+ "scene_present": true
139
+ }
scenes/isaac/rcd_lounge_4/20261114T065436Z/submitted_prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 30 minutes (1,800 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.
scenes/isaac/rcd_lounge_4/20261114T083703Z/evaluator_protocol.json ADDED
The diff for this file is too large to render. See raw diff
 
scenes/isaac/rcd_lounge_4/20261114T083703Z/generation.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/anchor_marked.png ADDED

Git LFS Details

  • SHA256: fc219ca78ee0730aaefac3d910420664fcf83c83747d04c6a8dfc97a3ec85405
  • Pointer size: 131 Bytes
  • Size of remote file: 272 kB
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/anchor_reference.png ADDED

Git LFS Details

  • SHA256: 02d769e29bd08628f82ebb850ed062b88897e9225aabc5134b12f5419dd2d7f8
  • Pointer size: 131 Bytes
  • Size of remote file: 273 kB
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/generation_spec.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "contract": "scenes-isaac",
3
+ "engine": "Isaac Sim",
4
+ "task": "Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.",
5
+ "inputs": [
6
+ "video.mp4",
7
+ "scale_anchor.json",
8
+ "anchor_reference.png",
9
+ "anchor_marked.png",
10
+ "self_check.py"
11
+ ],
12
+ "prompt": "prompt.txt",
13
+ "units": "metres, kilograms, seconds",
14
+ "up_axis": "+Z",
15
+ "entrypoint": "scene.usda",
16
+ "budget": {
17
+ "wall_clock_seconds": 900,
18
+ "scope": "per_scene",
19
+ "starts_at": "agent_input_access",
20
+ "includes_all_tool_and_validation_time": true,
21
+ "deadline_action": "stop and submit latest valid loadable output",
22
+ "enforcement": "Docker generation runner terminates the container at the wall-clock deadline"
23
+ },
24
+ "requirement": "Sim-ready: ready for interactive robot learning, with no manual repair.",
25
+ "sim_ready": {
26
+ "loads_headless_offline": true,
27
+ "resets_to_rest": true,
28
+ "collision_matches_visible_geometry": true,
29
+ "rigid_things": "rigid bodies; free objects movable, articulated parts on limited joints",
30
+ "soft_things": [
31
+ "cloth",
32
+ "bedding",
33
+ "curtains",
34
+ "rugs",
35
+ "throws",
36
+ "cushions",
37
+ "pillows"
38
+ ],
39
+ "soft_mechanism": "native deformable",
40
+ "stable_under_sustained_interaction": true
41
+ },
42
+ "visual_objective": "Maximize visual quality, realism and fidelity to the source video, keeping every sim-ready requirement.",
43
+ "refinement": {
44
+ "minimum_completed_rounds": 3,
45
+ "minimum_reference_views": 6,
46
+ "minimum_inspected_video_frames": 12,
47
+ "final_validation_seconds": 90,
48
+ "outside_data_allowed": false
49
+ },
50
+ "label": "free text saying plainly what the object is; one entry per physical instance",
51
+ "interaction_points": [
52
+ "handle",
53
+ "grasp",
54
+ "spout",
55
+ "material_corners"
56
+ ],
57
+ "construction_rules": {
58
+ "max_initial_penetration_m": 0.005,
59
+ "max_penetration_duration_s": 0.02,
60
+ "max_passive_speed_mps": 5.0,
61
+ "articulated_parts_clear_of_carcass": true,
62
+ "swing_paths_clear": true,
63
+ "timestep": "agent's choice"
64
+ },
65
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check"
66
+ }
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 15 minutes (900 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/scale_anchor.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "object_id": "scale_anchor",
3
+ "description": "The couch (2) marked in red in anchor_marked.png.",
4
+ "reference_image": "anchor_reference.png",
5
+ "marked_image": "anchor_marked.png",
6
+ "region_xyxy_pixels": [
7
+ 56.39891520732052,
8
+ 6.302853824313528,
9
+ 324.97904181777216,
10
+ 223.66539350903474
11
+ ],
12
+ "region_type": "projected_3d_box_envelope_not_segmentation",
13
+ "size_local_xyz_m": [
14
+ 1.256,
15
+ 0.838,
16
+ 0.859
17
+ ],
18
+ "axis_description": [
19
+ "horizontal length",
20
+ "horizontal width",
21
+ "vertical height"
22
+ ],
23
+ "measurement_type": "full extents of an oriented box around the object",
24
+ "uncertainty_m": null,
25
+ "uncertainty_note": "No uncertainty is supplied for these dimensions."
26
+ }
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/sdk_smoke.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Verify Isaac Sim headless RTX rendering and native rigid-body stepping."""
2
+ from isaacsim import SimulationApp
3
+ app = SimulationApp({'headless':True,'renderer':'PathTracing','samples_per_pixel_per_frame':32,
4
+ 'width':512,'height':384,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
5
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
6
+ import json
7
+ from pathlib import Path
8
+ import numpy as np
9
+ from PIL import Image
10
+ import omni.usd
11
+ import omni.replicator.core as rep
12
+ rep.orchestrator.set_capture_on_play(False)
13
+ from pxr import UsdGeom,UsdLux,UsdShade,UsdPhysics,Gf,Sdf
14
+ stage=omni.usd.get_context().get_stage()
15
+ UsdGeom.SetStageUpAxis(stage,UsdGeom.Tokens.z);UsdGeom.SetStageMetersPerUnit(stage,1.)
16
+ UsdGeom.Xform.Define(stage,'/World');stage.SetDefaultPrim(stage.GetPrimAtPath('/World'))
17
+ def box(path,pos,scale,color,dynamic=False):
18
+ prim=UsdGeom.Cube.Define(stage,path);prim.CreateSizeAttr(1.)
19
+ xf=UsdGeom.Xformable(prim);xf.AddTranslateOp().Set(Gf.Vec3d(*pos));xf.AddScaleOp().Set(Gf.Vec3f(*scale))
20
+ UsdPhysics.CollisionAPI.Apply(prim.GetPrim())
21
+ if dynamic:
22
+ UsdPhysics.RigidBodyAPI.Apply(prim.GetPrim());UsdPhysics.MassAPI.Apply(prim.GetPrim()).CreateMassAttr(1.)
23
+ mat=UsdShade.Material.Define(stage,path+'_Material');shader=UsdShade.Shader.Define(stage,path+'_Material/Shader');shader.CreateIdAttr('UsdPreviewSurface')
24
+ shader.CreateInput('diffuseColor',Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*color));shader.CreateInput('roughness',Sdf.ValueTypeNames.Float).Set(.28)
25
+ mat.CreateSurfaceOutput().ConnectToSource(shader.ConnectableAPI(),'surface');UsdShade.MaterialBindingAPI.Apply(prim.GetPrim()).Bind(mat)
26
+ return prim
27
+ box('/World/Floor',(0,0,-.05),(5,5,.1),(.6,.6,.6));box('/World/Cube',(0,0,1.),(.3,.3,.3),(.08,.3,.8),True)
28
+ light=UsdLux.DomeLight.Define(stage,'/World/Dome');light.CreateIntensityAttr(500.)
29
+ light=UsdLux.RectLight.Define(stage,'/World/Key');light.CreateIntensityAttr(2500.);light.CreateWidthAttr(3.);light.CreateHeightAttr(3.);UsdGeom.Xformable(light).AddTranslateOp().Set(Gf.Vec3d(0,0,3))
30
+ scene=UsdPhysics.Scene.Define(stage,'/World/PhysicsScene');scene.CreateGravityDirectionAttr(Gf.Vec3f(0,0,-1));scene.CreateGravityMagnitudeAttr(9.81)
31
+ camera=rep.create.camera(position=(2.4,-3,2),look_at=(0,0,.3));product=rep.create.render_product(camera,(512,384));rgb=rep.AnnotatorRegistry.get_annotator('rgb');rgb.attach([product])
32
+ from isaacsim.core.api import World
33
+ world=World(physics_dt=1/60,rendering_dt=1/60,stage_units_in_meters=1.)
34
+ world.reset()
35
+ for _ in range(30):world.step(render=True)
36
+ for attempt in range(30):
37
+ rep.orchestrator.step(rt_subframes=8)
38
+ app.update()
39
+ image=np.asarray(rgb.get_data())
40
+ print('CAPTURE_ATTEMPT',attempt,'SHAPE',image.shape,flush=True)
41
+ if image.ndim==3 and image.shape[0]>0:break
42
+ print('RGB_SHAPE',image.shape,flush=True)
43
+ assert image.ndim==3,'RTX annotator did not return a frame'
44
+ position=UsdGeom.Xformable(stage.GetPrimAtPath('/World/Cube')).ComputeLocalToWorldTransform(0).ExtractTranslation()
45
+ print('CUBE_POSITION',list(position),flush=True)
46
+ out=Path('/work/output');out.mkdir(exist_ok=True)
47
+ Image.fromarray(image[:,:,:3]).save(out/'smoke.png');stage.GetRootLayer().Export(str(out/'sdk_example.usda'))
48
+ (out/'smoke.json').write_text(json.dumps({'renderer':'Isaac Sim RTX PathTracing','shape':list(image.shape),'cube_final_position':list(position),'initial_height_m':1.,'success':True}))
49
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/self_check.py ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Passive physics of a USD scene in Isaac Sim, against the scene requirements (no penetration beyond 5 mm for more than
2
+ 0.02 s, no body faster than 5 m/s, finite state).
3
+
4
+ Runs inside the Isaac Sim container. Reads --scene (default /work/scene, read-only) and the optional
5
+ --out/request.json (object ids to drop); writes --out/physics.json (default /work/physics). Without a request it is
6
+ the self-check: no drops, plus a check that every object has a non-empty label.
7
+
8
+ - Speed: a 3 s passive rollout with no rendering and no state reads, timed around world.step only
9
+ (simulated seconds / wall seconds).
10
+ - Passive rollout: 3 s from the authored state. Penetration comes from the PhysX tensor contact view
11
+ (GPU contact reports return nothing under GPU dynamics); speed and poses from the RigidPrim view.
12
+ - Drops: the requested objects with free (unjointed) dynamic bodies, raised 0.05 m
13
+ and released for 2 s, each from a fresh reset.
14
+ The limits are the same as for MuJoCo scenes.
15
+ """
16
+ from isaacsim import SimulationApp
17
+ app=SimulationApp({'headless':True,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
18
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
19
+ import argparse
20
+ import json
21
+ from pathlib import Path
22
+ import time
23
+ import carb
24
+ import numpy as np
25
+ import omni.usd
26
+ from pxr import UsdPhysics,PhysxSchema
27
+
28
+ LIMIT_M,LIMIT_S,SPEED=.005,.02,5.
29
+ parser=argparse.ArgumentParser();parser.add_argument('--scene',default='/work/scene');parser.add_argument('--out',default='/work/physics')
30
+ args,_=parser.parse_known_args()
31
+ out=Path(args.out);out.mkdir(parents=True,exist_ok=True)
32
+ request=json.loads((out/'request.json').read_text()) if (out/'request.json').is_file() else {'object_ids':[]}
33
+ result={'status':'failed','engine':'Isaac Sim PhysX (GPU dynamics)','no_vlm':True,'provisional':True,
34
+ 'thresholds':{'penetration_m':LIMIT_M,'penetration_duration_s':LIMIT_S,'linear_speed_mps':SPEED},
35
+ 'thresholds_calibrated_across_engines':False}
36
+
37
+
38
+ def contact_filters(stage,paths):
39
+ """GPU contact views filter on rigid bodies or standalone colliders only, so a collider inside a
40
+ rigid body is replaced by that body (unsupported filters are dropped with only a warning)."""
41
+ out=[]
42
+ for path in paths:
43
+ if any('deformable' in str(a).lower() for a in stage.GetPrimAtPath(path).GetAppliedSchemas()):
44
+ continue # deformable meshes cannot be contact filters; their contacts are not measured
45
+ prim=stage.GetPrimAtPath(path);owner=None
46
+ while prim and not prim.IsPseudoRoot():
47
+ if prim.HasAPI(UsdPhysics.RigidBodyAPI):owner=str(prim.GetPath())
48
+ prim=prim.GetParent()
49
+ target=owner or path
50
+ if target not in out:out.append(target)
51
+ return out
52
+
53
+
54
+ def array(value):
55
+ return np.asarray(value.cpu() if hasattr(value,'cpu') else value,dtype=float)
56
+
57
+
58
+ try:
59
+ root=Path(args.scene)
60
+ scene=next(root/n for n in ['scene.usda','scene.usd'] if (root/n).is_file())
61
+ omni.usd.get_context().open_stage(str(scene));stage=omni.usd.get_context().get_stage()
62
+ for _ in range(8):app.update()
63
+
64
+ def dynamic(prim):
65
+ api=UsdPhysics.RigidBodyAPI(prim)
66
+ if api.GetRigidBodyEnabledAttr().Get() not in (None,True) or api.GetKinematicEnabledAttr().Get():return False
67
+ parent=prim.GetParent()
68
+ while parent and not parent.IsPseudoRoot():
69
+ if parent.HasAPI(UsdPhysics.RigidBodyAPI):return False
70
+ parent=parent.GetParent()
71
+ return True
72
+ rigid=[p for p in stage.Traverse() if p.HasAPI(UsdPhysics.RigidBodyAPI)]
73
+ bodies=[p for p in rigid if dynamic(p)]
74
+ paths=[str(p.GetPath()) for p in bodies]
75
+ colliders=[str(p.GetPath()) for p in stage.Traverse() if p.HasAPI(UsdPhysics.CollisionAPI)]
76
+ jointed=set()
77
+ for prim in stage.Traverse():
78
+ if prim.IsA(UsdPhysics.Joint):
79
+ joint=UsdPhysics.Joint(prim)
80
+ for rel in (joint.GetBody0Rel(),joint.GetBody1Rel()):
81
+ jointed.update(str(t) for t in rel.GetTargets())
82
+ for prim in bodies:PhysxSchema.PhysxContactReportAPI.Apply(prim).CreateThresholdAttr().Set(0.)
83
+ scenes=[p for p in stage.Traverse() if p.IsA(UsdPhysics.Scene)]
84
+ authored=PhysxSchema.PhysxSceneAPI(scenes[0]).GetTimeStepsPerSecondAttr().Get() if scenes else None
85
+ rate=float(authored or 60)
86
+ result.update(dynamic_bodies=len(paths),colliders=len(colliders),
87
+ excluded_bodies=[str(p.GetPath()) for p in rigid if not dynamic(p)],
88
+ timestep_s=1/rate,timestep_source='authored PhysxSceneAPI timeStepsPerSecond' if authored else 'default 60 Hz')
89
+ if not paths:raise ValueError('No dynamic rigid bodies to simulate')
90
+ filters=contact_filters(stage,colliders)
91
+ result['contact_filters']=len(filters)
92
+
93
+ carb.settings.get_settings().set_bool('/physics/disableContactProcessing',False)
94
+ from isaacsim.core.api import World
95
+ from isaacsim.core.prims import RigidPrim
96
+ from isaacsim.core.simulation_manager import SimulationManager
97
+ world=World(physics_dt=1/rate,rendering_dt=1/rate,stage_units_in_meters=1.,
98
+ physics_prim_path=str(scenes[0].GetPath()) if scenes else '/World/AddedPhysicsScene',
99
+ set_defaults=False,backend='torch',device='cuda:0')
100
+ context=world.get_physics_context()
101
+ context.enable_gpu_dynamics(True);context.set_broadphase_type('GPU')
102
+ view=RigidPrim(paths,name='scene_bodies');world.scene.add(view)
103
+ steps=lambda seconds:int(np.ceil(seconds*rate))
104
+
105
+ def measure_rtf(dt,repeats=3):
106
+ """Simulated seconds / wall seconds of world.step(render=False) for an unread passive rollout from the authored
107
+ state, median over `repeats` after 10 warm-up steps."""
108
+ world.set_simulation_dt(physics_dt=dt,rendering_dt=dt)
109
+ world.reset()
110
+ for _ in range(10):world.step(render=False)
111
+ count=int(np.ceil(3/dt));samples=[];finite=True
112
+ for _ in range(repeats):
113
+ world.reset();start=time.perf_counter()
114
+ for _ in range(count):world.step(render=False)
115
+ samples.append(count*dt/(time.perf_counter()-start))
116
+ positions,_=view.get_world_poses()
117
+ finite&=bool(np.isfinite(array(positions)).all() and np.isfinite(array(view.get_velocities())).all())
118
+ return {'rtf':float(np.median(samples)),'rtf_samples':samples,'timestep_s':dt,'simulated_seconds':count*dt,
119
+ 'steps':count,'repeats':repeats,'finite':finite,
120
+ 'definition':'simulated seconds / wall seconds of world.step(render=False) only, median of repeats'}
121
+
122
+ result['realtime']={**measure_rtf(1/rate),'scope':'3 s passive rollout from the authored state, authored timestep'}
123
+ print('PHYSICS_STAGE realtime',result['realtime']['rtf'],flush=True)
124
+
125
+ def reset():
126
+ world.reset()
127
+ contacts=SimulationManager.get_physics_sim_view().create_rigid_contact_view(
128
+ paths,filter_patterns=[filters]*len(paths),max_contact_data_count=200000)
129
+ if contacts is None:raise RuntimeError('PhysX rigid contact view could not be created')
130
+ return contacts
131
+
132
+ def rollout(seconds,contacts,motion=None):
133
+ """One rollout; `motion` names the file the logged poses are written to."""
134
+ report={'finite':True,'peak_penetration_m':0.,'penetration_time_s':0.,'peak_linear_speed_mps':0.}
135
+ every=max(1,round(rate/10));frames=[]
136
+ for step in range(steps(seconds)):
137
+ world.step(render=False)
138
+ positions,orientations=view.get_world_poses();velocities=array(view.get_velocities())
139
+ if step%every==0:frames.append(((step+1)/rate,array(positions),array(orientations)))
140
+ report['finite']&=bool(np.isfinite(array(positions)).all() and np.isfinite(velocities).all())
141
+ _,_,_,separation,counts,starts=contacts.get_contact_data(dt=1/rate)
142
+ total=int(array(counts).sum())
143
+ if 'initial_pairs' not in report:
144
+ # Pairs deeper than the limit after the first step, named for the self-check.
145
+ pairs=[];sep=array(separation);cnt=array(counts).astype(int);beg=array(starts).astype(int)
146
+ for i,j in zip(*np.nonzero(cnt)):
147
+ depth=-float(sep[beg[i,j]:beg[i,j]+cnt[i,j]].min())
148
+ if depth>LIMIT_M:pairs.append({'depth_m':depth,'bodies':[paths[i],filters[j]]})
149
+ report['initial_pairs']=sorted(pairs,key=lambda p:-p['depth_m'])[:20]
150
+ deepest=max(0.,-float(array(separation)[:total].min())) if total else 0.
151
+ report['peak_penetration_m']=max(report['peak_penetration_m'],deepest)
152
+ if deepest>LIMIT_M:report['penetration_time_s']+=1/rate
153
+ report['peak_linear_speed_mps']=max(report['peak_linear_speed_mps'],float(np.linalg.norm(velocities[:,:3],axis=1).max()))
154
+ if not report['finite']:break
155
+ report['valid']=bool(report['finite'] and report['penetration_time_s']<=LIMIT_S and report['peak_linear_speed_mps']<=SPEED)
156
+ report['thresholds']=result['thresholds']
157
+ if motion and request.get('record'):
158
+ (out/'motion').mkdir(exist_ok=True)
159
+ np.savez_compressed(out/'motion'/motion,paths=np.array(paths),t=np.array([f[0] for f in frames]),
160
+ positions=np.stack([f[1] for f in frames]),orientations=np.stack([f[2] for f in frames]))
161
+ report['motion']=f'motion/{motion}.npz'
162
+ return report
163
+
164
+ contacts=reset()
165
+ start_positions=array(view.get_world_poses()[0])
166
+ masses=array(view.get_masses())
167
+ direction,magnitude=context.get_gravity()
168
+ gravity=np.asarray(direction,dtype=float)*float(magnitude)
169
+ passive=rollout(3,contacts,'passive')
170
+ print('PHYSICS_STAGE passive',passive,flush=True)
171
+ end_positions=array(view.get_world_poses()[0])
172
+ moved=np.linalg.norm(end_positions-start_positions,axis=1)
173
+
174
+ manifest=json.loads((root/'objects.json').read_text())
175
+ manifest=manifest['objects'] if isinstance(manifest,dict) else manifest
176
+ result['manifest_errors']=[f"{o.get('id')}: label is missing or empty" for o in manifest if not str(o.get('label') or '').strip()]
177
+ result['manifest_errors']+=[f"{o.get('id')}: prim_path does not resolve" for o in manifest if not stage.GetPrimAtPath(o.get('prim_path','') or '/')]
178
+ def owned(obj):
179
+ prefix=str(obj.get('prim_path','')).rstrip('/')
180
+ return [i for i,p in enumerate(paths) if prefix and (p==prefix or p.startswith(prefix+'/'))]
181
+ drift=[{'id':o['id'],'dynamic_bodies':len(owned(o)),'center_drift_m':float(moved[owned(o)].max()),'mass_kg':float(masses[owned(o)].sum())}
182
+ for o in manifest if owned(o)]
183
+
184
+ requested=set(request['object_ids'])
185
+ eligible=[o for o in manifest if o['id'] in requested]
186
+ drops=[]
187
+ for obj in eligible:
188
+ free=[i for i in owned(obj) if paths[i] not in jointed]
189
+ if not free:continue
190
+ contacts=reset()
191
+ positions,orientations=view.get_world_poses()
192
+ raised=positions.clone() if hasattr(positions,'clone') else positions.copy()
193
+ raised[free,2]+=.05
194
+ view.set_world_poses(positions=raised,orientations=orientations)
195
+ applied=float((array(view.get_world_poses()[0])[free,2]-array(positions)[free,2]).min())
196
+ drops.append({'id':obj['id'],'input':'raise free bodies 0.05 m and release','bodies':[paths[i] for i in free],
197
+ 'lift_applied_m':applied,**rollout(2,contacts,f'drop_{len(drops)}')})
198
+ print('PHYSICS_STAGE drop',obj['id'],applied,drops[-1]['peak_linear_speed_mps'],flush=True)
199
+
200
+ # The common 2 ms step last, so the rollouts above run at the authored step.
201
+ result['realtime_common_step']={**measure_rtf(.002),'scope':'3 s passive rollout from the authored state at the common 2 ms step'}
202
+ print('PHYSICS_STAGE realtime_common_step',result['realtime_common_step']['rtf'],flush=True)
203
+ gravity_ok=bool(np.allclose(gravity,[0,0,-9.81],atol=.1))
204
+ mass_ok=bool(np.isfinite(masses).all() and (masses>=0).all())
205
+ result.update(status='completed',passive=passive,gravity_m_s2=gravity.tolist(),gravity_valid=gravity_ok,
206
+ mass_valid=mass_ok,center_drift=drift,drop_tests=drops,eligible_native_objects=len(eligible),
207
+ largest_body_displacements=[{'prim':paths[i],'displacement_m':float(moved[i])} for i in np.argsort(-moved)[:5]],
208
+ valid=bool(passive['valid'] and gravity_ok and mass_ok),
209
+ drop_scope='the requested objects with free (unjointed) dynamic bodies, in manifest order',
210
+ contact_scope=('contacts between each dynamic body and every collider (static, kinematic and dynamic; colliders '
211
+ 'inside a rigid body are filtered by that body, as GPU contact views require); '
212
+ 'static-static pairs are not simulated, as in MuJoCo world-body geoms'),
213
+ limitations=['the limits are the MuJoCo ones, not calibrated for PhysX',
214
+ 'kinematic, disabled and nested rigid bodies are not traced (see excluded_bodies)',
215
+ 'drift is per dynamic body under the object prim, not the object bounding box',
216
+ 'contacts with deformable bodies are not measured (GPU contact views cannot filter them)'])
217
+ except Exception as error:
218
+ result['error']=f'{type(error).__name__}: {error}'
219
+ finally:
220
+ (out/'physics.json').write_text(json.dumps(result,indent=2)+'\n')
221
+ if 'passive' in result:
222
+ passive=result['passive']
223
+ print('PASSIVE CHECK','PASSED' if passive['valid'] and not result.get('manifest_errors') else 'FAILED',
224
+ f"- {len(passive.get('initial_pairs',[]))} contact pairs > 5 mm, {passive['penetration_time_s']:.3f} s above 5 mm,"
225
+ f" peak speed {passive['peak_linear_speed_mps']:.2f} m/s, {len(result.get('manifest_errors',[]))} manifest errors;"
226
+ f" details in {out/'physics.json'}",flush=True)
227
+ else:print('CHECK FAILED:',result.get('error'),flush=True)
228
+ app.close()
scenes/isaac/rcd_lounge_4/20261114T083703Z/input/video.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:266549020e323df9619a00d353edef099d840ba582bfa732ccc7af90220e1fe4
3
+ size 16239519
scenes/isaac/rcd_lounge_4/20261114T083703Z/run.json ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "track": "scenes",
3
+ "engine": "isaac",
4
+ "item": "rcd_lounge_4",
5
+ "run_id": "20261114T083703Z",
6
+ "campaign": "jiarui-2-codebuddy",
7
+ "cli": "codebuddy",
8
+ "model": "Hunyuan 4",
9
+ "model_id": "hy4-preview",
10
+ "model_family": "hy4-preview",
11
+ "model_reasoning_effort": "default",
12
+ "budget_seconds": 900,
13
+ "told_budget_seconds": 900,
14
+ "stage": 1,
15
+ "continued_from": null,
16
+ "contract": {
17
+ "name": "scenes-isaac",
18
+ "sha256": "f1ab7ea88842502d37176b41d9f3bec64a5d9fac292ca3f16caf08da3b5db96d",
19
+ "track": "scenes",
20
+ "engine": "isaac",
21
+ "engine_version": "6.2.0",
22
+ "image": "simready-isaac-generation:latest",
23
+ "blender_version": "5.2.3",
24
+ "inputs": [
25
+ "video.mp4",
26
+ "scale_anchor.json",
27
+ "anchor_reference.png",
28
+ "anchor_marked.png",
29
+ "prompt.txt",
30
+ "generation_spec.json",
31
+ "self_check.py",
32
+ "sdk_smoke.py"
33
+ ],
34
+ "withheld": [
35
+ "groundtruth/ (objects, cameras, task catalogs, object dynamics, interaction test) and its source tag",
36
+ "provenance/ (source annotation, anchor selection)",
37
+ "the frozen protocol of the track and engine (data/benchmark/frozen/<track>/<engine>.json)",
38
+ "other runs"
39
+ ],
40
+ "outputs": [
41
+ "scene.usda",
42
+ "build_scene.py",
43
+ "objects.json",
44
+ "render_preview.py",
45
+ "preview.mp4",
46
+ "interaction.mp4",
47
+ "README.md",
48
+ "refinement_log.json"
49
+ ],
50
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check",
51
+ "budget": {
52
+ "told": "the first of the session's cumulative budgets, in wall-clock seconds from container start",
53
+ "resume": "each later budget resumes the same conversation for the rest of its time; each budget is its own run",
54
+ "enforcement": "the runner kills the container at the deadline; the output as it stands is the result"
55
+ },
56
+ "refinement": {
57
+ "minimum_completed_rounds": 3,
58
+ "minimum_reference_views": 6,
59
+ "minimum_inspected_video_frames": 12,
60
+ "final_validation_seconds": 90
61
+ },
62
+ "isolation": {
63
+ "container": "Docker: read-only input (and tools), writable output and scratch, temporary credentials; no repository, benchmark ground truth or evaluator files",
64
+ "network": "an internal network whose only exit is an HTTPS CONNECT proxy (runtime/egress_proxy.py)",
65
+ "egress_hosts": [
66
+ "api.anthropic.com",
67
+ "platform.claude.com",
68
+ "chatgpt.com",
69
+ "auth.openai.com",
70
+ "opencode.ai",
71
+ "mcp.exa.ai",
72
+ "search.parallel.ai"
73
+ ]
74
+ },
75
+ "protocol": {
76
+ "file": "data/benchmark/frozen/scenes/isaac.json",
77
+ "sha256": "df0fe5bb1c796474ad13cdba81b660a2946aa24297cb289f9581cfde7c035bb7"
78
+ }
79
+ },
80
+ "engine_version": "6.2.0",
81
+ "image": "simready-isaac-generation:latest",
82
+ "image_id": "sha256:8d9f9c934457aa3c8e91e8099e35db27b50db287d0eb76a647052c310e754d87",
83
+ "required_refinement_rounds": 3,
84
+ "gpu_uuid": "GPU-ab037b41-847d-64be-b2b8-ba682a201943",
85
+ "egress_mode": "internal_network",
86
+ "input_snapshot": "input",
87
+ "tools": null,
88
+ "evaluator_protocol": "evaluator_protocol.json",
89
+ "output": "output",
90
+ "log": "generation.jsonl",
91
+ "status": "budget_expired",
92
+ "agent_cli": {
93
+ "cli": "codebuddy",
94
+ "name": "CodeBuddy Code",
95
+ "binary": "a pinned codebuddy-code package and its node, mounted read-only (SIMREADY_CODEBUDDY_PREFIX, default the host's codebuddy install)",
96
+ "login": "the CodeBuddy login file, given to the CLI as a snapshot without its refresh token (the session cannot rotate it)",
97
+ "billing": "CodeBuddy account (the CLI reports no price)",
98
+ "tools": "every tool but WebFetch and the server-side media generators ImageGen, ImageEdit, VideoGen (--disallowedTools); WebSearch runs on CodeBuddy's side",
99
+ "hosts": [
100
+ "[redacted-host]",
101
+ "wb.[redacted]buddy.com"
102
+ ],
103
+ "version": "2.182.0",
104
+ "model": "hy4-preview",
105
+ "model_family": "hy4-preview"
106
+ },
107
+ "started_at": "2026-11-14T08:37:04.847669+00:00",
108
+ "model_answered": true,
109
+ "exit_code": 137,
110
+ "credential_refresh_persistence": "snapshot_without_refresh_token",
111
+ "finished_at": "2026-11-14T08:52:06.722414+00:00",
112
+ "cli_session_id": "01a1d934-0422-7fb9-b2eb-f34e0963922b",
113
+ "api_retries": 0,
114
+ "rate_limit_events": 0,
115
+ "stage_seconds": 901.916,
116
+ "elapsed_seconds": 901.916,
117
+ "cost": {
118
+ "status": "partial",
119
+ "source": "the CLI's running usage totals (usage_update events)",
120
+ "usd": null,
121
+ "usd_basis": "not reported (CodeBuddy account billing)",
122
+ "credits": 0,
123
+ "credit_status": "known",
124
+ "turns": null,
125
+ "api_seconds": null,
126
+ "tokens": {
127
+ "input": 80193,
128
+ "cache_read": 405760,
129
+ "cache_write": 0,
130
+ "output": 9093,
131
+ "reasoning": null,
132
+ "total": 495046
133
+ },
134
+ "wall_seconds": 901.916,
135
+ "budget_seconds": 900,
136
+ "stopped_at_budget": true
137
+ },
138
+ "scene_present": false
139
+ }
scenes/isaac/rcd_lounge_4/20261114T083703Z/submitted_prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 15 minutes (900 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/evaluator_protocol.json ADDED
The diff for this file is too large to render. See raw diff
 
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/generation.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5629d28b7043f51bd5a24aca4fd3294610122335f1c4f211d8d0f2aa039de293
3
+ size 13844222
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/anchor_marked.png ADDED

Git LFS Details

  • SHA256: 4411a96d4851cec35f3ccfdce323859882f1d6be582d2762bb23f67063d9c8c6
  • Pointer size: 131 Bytes
  • Size of remote file: 184 kB
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/anchor_reference.png ADDED

Git LFS Details

  • SHA256: 1cd422e428a78145db365585e34c7229d2c6f9c08d09c875db97c18507d1d133
  • Pointer size: 131 Bytes
  • Size of remote file: 185 kB
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/generation_spec.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "contract": "scenes-isaac",
3
+ "engine": "Isaac Sim",
4
+ "task": "Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.",
5
+ "inputs": [
6
+ "video.mp4",
7
+ "scale_anchor.json",
8
+ "anchor_reference.png",
9
+ "anchor_marked.png",
10
+ "self_check.py"
11
+ ],
12
+ "prompt": "prompt.txt",
13
+ "units": "metres, kilograms, seconds",
14
+ "up_axis": "+Z",
15
+ "entrypoint": "scene.usda",
16
+ "budget": {
17
+ "wall_clock_seconds": 1800,
18
+ "scope": "per_scene",
19
+ "starts_at": "agent_input_access",
20
+ "includes_all_tool_and_validation_time": true,
21
+ "deadline_action": "stop and submit latest valid loadable output",
22
+ "enforcement": "Docker generation runner terminates the container at the wall-clock deadline"
23
+ },
24
+ "requirement": "Sim-ready: ready for interactive robot learning, with no manual repair.",
25
+ "sim_ready": {
26
+ "loads_headless_offline": true,
27
+ "resets_to_rest": true,
28
+ "collision_matches_visible_geometry": true,
29
+ "rigid_things": "rigid bodies; free objects movable, articulated parts on limited joints",
30
+ "soft_things": [
31
+ "cloth",
32
+ "bedding",
33
+ "curtains",
34
+ "rugs",
35
+ "throws",
36
+ "cushions",
37
+ "pillows"
38
+ ],
39
+ "soft_mechanism": "native deformable",
40
+ "stable_under_sustained_interaction": true
41
+ },
42
+ "visual_objective": "Maximize visual quality, realism and fidelity to the source video, keeping every sim-ready requirement.",
43
+ "refinement": {
44
+ "minimum_completed_rounds": 3,
45
+ "minimum_reference_views": 6,
46
+ "minimum_inspected_video_frames": 12,
47
+ "final_validation_seconds": 90,
48
+ "outside_data_allowed": false
49
+ },
50
+ "label": "free text saying plainly what the object is; one entry per physical instance",
51
+ "interaction_points": [
52
+ "handle",
53
+ "grasp",
54
+ "spout",
55
+ "material_corners"
56
+ ],
57
+ "construction_rules": {
58
+ "max_initial_penetration_m": 0.005,
59
+ "max_penetration_duration_s": 0.02,
60
+ "max_passive_speed_mps": 5.0,
61
+ "articulated_parts_clear_of_carcass": true,
62
+ "swing_paths_clear": true,
63
+ "timestep": "agent's choice"
64
+ },
65
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check"
66
+ }
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 30 minutes (1,800 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/scale_anchor.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "object_id": "scale_anchor",
3
+ "description": "The white trash bin with a plastic liner, at the red box in anchor_marked.png; the box is drawn slightly to the left of the bin.",
4
+ "reference_image": "anchor_reference.png",
5
+ "marked_image": "anchor_marked.png",
6
+ "region_xyxy_pixels": [
7
+ 92.3369602003388,
8
+ 396.77753645958376,
9
+ 225.4898154070338,
10
+ 627.3115717283308
11
+ ],
12
+ "region_type": "projected_3d_box_envelope_not_segmentation",
13
+ "size_local_xyz_m": [
14
+ 0.418,
15
+ 0.294,
16
+ 0.483
17
+ ],
18
+ "axis_description": [
19
+ "horizontal length",
20
+ "horizontal width",
21
+ "vertical height"
22
+ ],
23
+ "measurement_type": "full extents of an oriented box around the object",
24
+ "uncertainty_m": null,
25
+ "uncertainty_note": "No uncertainty is supplied for these dimensions."
26
+ }
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/sdk_smoke.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Verify Isaac Sim headless RTX rendering and native rigid-body stepping."""
2
+ from isaacsim import SimulationApp
3
+ app = SimulationApp({'headless':True,'renderer':'PathTracing','samples_per_pixel_per_frame':32,
4
+ 'width':512,'height':384,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
5
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
6
+ import json
7
+ from pathlib import Path
8
+ import numpy as np
9
+ from PIL import Image
10
+ import omni.usd
11
+ import omni.replicator.core as rep
12
+ rep.orchestrator.set_capture_on_play(False)
13
+ from pxr import UsdGeom,UsdLux,UsdShade,UsdPhysics,Gf,Sdf
14
+ stage=omni.usd.get_context().get_stage()
15
+ UsdGeom.SetStageUpAxis(stage,UsdGeom.Tokens.z);UsdGeom.SetStageMetersPerUnit(stage,1.)
16
+ UsdGeom.Xform.Define(stage,'/World');stage.SetDefaultPrim(stage.GetPrimAtPath('/World'))
17
+ def box(path,pos,scale,color,dynamic=False):
18
+ prim=UsdGeom.Cube.Define(stage,path);prim.CreateSizeAttr(1.)
19
+ xf=UsdGeom.Xformable(prim);xf.AddTranslateOp().Set(Gf.Vec3d(*pos));xf.AddScaleOp().Set(Gf.Vec3f(*scale))
20
+ UsdPhysics.CollisionAPI.Apply(prim.GetPrim())
21
+ if dynamic:
22
+ UsdPhysics.RigidBodyAPI.Apply(prim.GetPrim());UsdPhysics.MassAPI.Apply(prim.GetPrim()).CreateMassAttr(1.)
23
+ mat=UsdShade.Material.Define(stage,path+'_Material');shader=UsdShade.Shader.Define(stage,path+'_Material/Shader');shader.CreateIdAttr('UsdPreviewSurface')
24
+ shader.CreateInput('diffuseColor',Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*color));shader.CreateInput('roughness',Sdf.ValueTypeNames.Float).Set(.28)
25
+ mat.CreateSurfaceOutput().ConnectToSource(shader.ConnectableAPI(),'surface');UsdShade.MaterialBindingAPI.Apply(prim.GetPrim()).Bind(mat)
26
+ return prim
27
+ box('/World/Floor',(0,0,-.05),(5,5,.1),(.6,.6,.6));box('/World/Cube',(0,0,1.),(.3,.3,.3),(.08,.3,.8),True)
28
+ light=UsdLux.DomeLight.Define(stage,'/World/Dome');light.CreateIntensityAttr(500.)
29
+ light=UsdLux.RectLight.Define(stage,'/World/Key');light.CreateIntensityAttr(2500.);light.CreateWidthAttr(3.);light.CreateHeightAttr(3.);UsdGeom.Xformable(light).AddTranslateOp().Set(Gf.Vec3d(0,0,3))
30
+ scene=UsdPhysics.Scene.Define(stage,'/World/PhysicsScene');scene.CreateGravityDirectionAttr(Gf.Vec3f(0,0,-1));scene.CreateGravityMagnitudeAttr(9.81)
31
+ camera=rep.create.camera(position=(2.4,-3,2),look_at=(0,0,.3));product=rep.create.render_product(camera,(512,384));rgb=rep.AnnotatorRegistry.get_annotator('rgb');rgb.attach([product])
32
+ from isaacsim.core.api import World
33
+ world=World(physics_dt=1/60,rendering_dt=1/60,stage_units_in_meters=1.)
34
+ world.reset()
35
+ for _ in range(30):world.step(render=True)
36
+ for attempt in range(30):
37
+ rep.orchestrator.step(rt_subframes=8)
38
+ app.update()
39
+ image=np.asarray(rgb.get_data())
40
+ print('CAPTURE_ATTEMPT',attempt,'SHAPE',image.shape,flush=True)
41
+ if image.ndim==3 and image.shape[0]>0:break
42
+ print('RGB_SHAPE',image.shape,flush=True)
43
+ assert image.ndim==3,'RTX annotator did not return a frame'
44
+ position=UsdGeom.Xformable(stage.GetPrimAtPath('/World/Cube')).ComputeLocalToWorldTransform(0).ExtractTranslation()
45
+ print('CUBE_POSITION',list(position),flush=True)
46
+ out=Path('/work/output');out.mkdir(exist_ok=True)
47
+ Image.fromarray(image[:,:,:3]).save(out/'smoke.png');stage.GetRootLayer().Export(str(out/'sdk_example.usda'))
48
+ (out/'smoke.json').write_text(json.dumps({'renderer':'Isaac Sim RTX PathTracing','shape':list(image.shape),'cube_final_position':list(position),'initial_height_m':1.,'success':True}))
49
+ app.close()
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/self_check.py ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Passive physics of a USD scene in Isaac Sim, against the scene requirements (no penetration beyond 5 mm for more than
2
+ 0.02 s, no body faster than 5 m/s, finite state).
3
+
4
+ Runs inside the Isaac Sim container. Reads --scene (default /work/scene, read-only) and the optional
5
+ --out/request.json (object ids to drop); writes --out/physics.json (default /work/physics). Without a request it is
6
+ the self-check: no drops, plus a check that every object has a non-empty label.
7
+
8
+ - Speed: a 3 s passive rollout with no rendering and no state reads, timed around world.step only
9
+ (simulated seconds / wall seconds).
10
+ - Passive rollout: 3 s from the authored state. Penetration comes from the PhysX tensor contact view
11
+ (GPU contact reports return nothing under GPU dynamics); speed and poses from the RigidPrim view.
12
+ - Drops: the requested objects with free (unjointed) dynamic bodies, raised 0.05 m
13
+ and released for 2 s, each from a fresh reset.
14
+ The limits are the same as for MuJoCo scenes.
15
+ """
16
+ from isaacsim import SimulationApp
17
+ app=SimulationApp({'headless':True,'multi_gpu':False,'active_gpu':0,'physics_gpu':0,
18
+ 'enable_crashreporter':False,'limit_cpu_threads':8})
19
+ import argparse
20
+ import json
21
+ from pathlib import Path
22
+ import time
23
+ import carb
24
+ import numpy as np
25
+ import omni.usd
26
+ from pxr import UsdPhysics,PhysxSchema
27
+
28
+ LIMIT_M,LIMIT_S,SPEED=.005,.02,5.
29
+ parser=argparse.ArgumentParser();parser.add_argument('--scene',default='/work/scene');parser.add_argument('--out',default='/work/physics')
30
+ args,_=parser.parse_known_args()
31
+ out=Path(args.out);out.mkdir(parents=True,exist_ok=True)
32
+ request=json.loads((out/'request.json').read_text()) if (out/'request.json').is_file() else {'object_ids':[]}
33
+ result={'status':'failed','engine':'Isaac Sim PhysX (GPU dynamics)','no_vlm':True,'provisional':True,
34
+ 'thresholds':{'penetration_m':LIMIT_M,'penetration_duration_s':LIMIT_S,'linear_speed_mps':SPEED},
35
+ 'thresholds_calibrated_across_engines':False}
36
+
37
+
38
+ def contact_filters(stage,paths):
39
+ """GPU contact views filter on rigid bodies or standalone colliders only, so a collider inside a
40
+ rigid body is replaced by that body (unsupported filters are dropped with only a warning)."""
41
+ out=[]
42
+ for path in paths:
43
+ if any('deformable' in str(a).lower() for a in stage.GetPrimAtPath(path).GetAppliedSchemas()):
44
+ continue # deformable meshes cannot be contact filters; their contacts are not measured
45
+ prim=stage.GetPrimAtPath(path);owner=None
46
+ while prim and not prim.IsPseudoRoot():
47
+ if prim.HasAPI(UsdPhysics.RigidBodyAPI):owner=str(prim.GetPath())
48
+ prim=prim.GetParent()
49
+ target=owner or path
50
+ if target not in out:out.append(target)
51
+ return out
52
+
53
+
54
+ def array(value):
55
+ return np.asarray(value.cpu() if hasattr(value,'cpu') else value,dtype=float)
56
+
57
+
58
+ try:
59
+ root=Path(args.scene)
60
+ scene=next(root/n for n in ['scene.usda','scene.usd'] if (root/n).is_file())
61
+ omni.usd.get_context().open_stage(str(scene));stage=omni.usd.get_context().get_stage()
62
+ for _ in range(8):app.update()
63
+
64
+ def dynamic(prim):
65
+ api=UsdPhysics.RigidBodyAPI(prim)
66
+ if api.GetRigidBodyEnabledAttr().Get() not in (None,True) or api.GetKinematicEnabledAttr().Get():return False
67
+ parent=prim.GetParent()
68
+ while parent and not parent.IsPseudoRoot():
69
+ if parent.HasAPI(UsdPhysics.RigidBodyAPI):return False
70
+ parent=parent.GetParent()
71
+ return True
72
+ rigid=[p for p in stage.Traverse() if p.HasAPI(UsdPhysics.RigidBodyAPI)]
73
+ bodies=[p for p in rigid if dynamic(p)]
74
+ paths=[str(p.GetPath()) for p in bodies]
75
+ colliders=[str(p.GetPath()) for p in stage.Traverse() if p.HasAPI(UsdPhysics.CollisionAPI)]
76
+ jointed=set()
77
+ for prim in stage.Traverse():
78
+ if prim.IsA(UsdPhysics.Joint):
79
+ joint=UsdPhysics.Joint(prim)
80
+ for rel in (joint.GetBody0Rel(),joint.GetBody1Rel()):
81
+ jointed.update(str(t) for t in rel.GetTargets())
82
+ for prim in bodies:PhysxSchema.PhysxContactReportAPI.Apply(prim).CreateThresholdAttr().Set(0.)
83
+ scenes=[p for p in stage.Traverse() if p.IsA(UsdPhysics.Scene)]
84
+ authored=PhysxSchema.PhysxSceneAPI(scenes[0]).GetTimeStepsPerSecondAttr().Get() if scenes else None
85
+ rate=float(authored or 60)
86
+ result.update(dynamic_bodies=len(paths),colliders=len(colliders),
87
+ excluded_bodies=[str(p.GetPath()) for p in rigid if not dynamic(p)],
88
+ timestep_s=1/rate,timestep_source='authored PhysxSceneAPI timeStepsPerSecond' if authored else 'default 60 Hz')
89
+ if not paths:raise ValueError('No dynamic rigid bodies to simulate')
90
+ filters=contact_filters(stage,colliders)
91
+ result['contact_filters']=len(filters)
92
+
93
+ carb.settings.get_settings().set_bool('/physics/disableContactProcessing',False)
94
+ from isaacsim.core.api import World
95
+ from isaacsim.core.prims import RigidPrim
96
+ from isaacsim.core.simulation_manager import SimulationManager
97
+ world=World(physics_dt=1/rate,rendering_dt=1/rate,stage_units_in_meters=1.,
98
+ physics_prim_path=str(scenes[0].GetPath()) if scenes else '/World/AddedPhysicsScene',
99
+ set_defaults=False,backend='torch',device='cuda:0')
100
+ context=world.get_physics_context()
101
+ context.enable_gpu_dynamics(True);context.set_broadphase_type('GPU')
102
+ view=RigidPrim(paths,name='scene_bodies');world.scene.add(view)
103
+ steps=lambda seconds:int(np.ceil(seconds*rate))
104
+
105
+ def measure_rtf(dt,repeats=3):
106
+ """Simulated seconds / wall seconds of world.step(render=False) for an unread passive rollout from the authored
107
+ state, median over `repeats` after 10 warm-up steps."""
108
+ world.set_simulation_dt(physics_dt=dt,rendering_dt=dt)
109
+ world.reset()
110
+ for _ in range(10):world.step(render=False)
111
+ count=int(np.ceil(3/dt));samples=[];finite=True
112
+ for _ in range(repeats):
113
+ world.reset();start=time.perf_counter()
114
+ for _ in range(count):world.step(render=False)
115
+ samples.append(count*dt/(time.perf_counter()-start))
116
+ positions,_=view.get_world_poses()
117
+ finite&=bool(np.isfinite(array(positions)).all() and np.isfinite(array(view.get_velocities())).all())
118
+ return {'rtf':float(np.median(samples)),'rtf_samples':samples,'timestep_s':dt,'simulated_seconds':count*dt,
119
+ 'steps':count,'repeats':repeats,'finite':finite,
120
+ 'definition':'simulated seconds / wall seconds of world.step(render=False) only, median of repeats'}
121
+
122
+ result['realtime']={**measure_rtf(1/rate),'scope':'3 s passive rollout from the authored state, authored timestep'}
123
+ print('PHYSICS_STAGE realtime',result['realtime']['rtf'],flush=True)
124
+
125
+ def reset():
126
+ world.reset()
127
+ contacts=SimulationManager.get_physics_sim_view().create_rigid_contact_view(
128
+ paths,filter_patterns=[filters]*len(paths),max_contact_data_count=200000)
129
+ if contacts is None:raise RuntimeError('PhysX rigid contact view could not be created')
130
+ return contacts
131
+
132
+ def rollout(seconds,contacts,motion=None):
133
+ """One rollout; `motion` names the file the logged poses are written to."""
134
+ report={'finite':True,'peak_penetration_m':0.,'penetration_time_s':0.,'peak_linear_speed_mps':0.}
135
+ every=max(1,round(rate/10));frames=[]
136
+ for step in range(steps(seconds)):
137
+ world.step(render=False)
138
+ positions,orientations=view.get_world_poses();velocities=array(view.get_velocities())
139
+ if step%every==0:frames.append(((step+1)/rate,array(positions),array(orientations)))
140
+ report['finite']&=bool(np.isfinite(array(positions)).all() and np.isfinite(velocities).all())
141
+ _,_,_,separation,counts,starts=contacts.get_contact_data(dt=1/rate)
142
+ total=int(array(counts).sum())
143
+ if 'initial_pairs' not in report:
144
+ # Pairs deeper than the limit after the first step, named for the self-check.
145
+ pairs=[];sep=array(separation);cnt=array(counts).astype(int);beg=array(starts).astype(int)
146
+ for i,j in zip(*np.nonzero(cnt)):
147
+ depth=-float(sep[beg[i,j]:beg[i,j]+cnt[i,j]].min())
148
+ if depth>LIMIT_M:pairs.append({'depth_m':depth,'bodies':[paths[i],filters[j]]})
149
+ report['initial_pairs']=sorted(pairs,key=lambda p:-p['depth_m'])[:20]
150
+ deepest=max(0.,-float(array(separation)[:total].min())) if total else 0.
151
+ report['peak_penetration_m']=max(report['peak_penetration_m'],deepest)
152
+ if deepest>LIMIT_M:report['penetration_time_s']+=1/rate
153
+ report['peak_linear_speed_mps']=max(report['peak_linear_speed_mps'],float(np.linalg.norm(velocities[:,:3],axis=1).max()))
154
+ if not report['finite']:break
155
+ report['valid']=bool(report['finite'] and report['penetration_time_s']<=LIMIT_S and report['peak_linear_speed_mps']<=SPEED)
156
+ report['thresholds']=result['thresholds']
157
+ if motion and request.get('record'):
158
+ (out/'motion').mkdir(exist_ok=True)
159
+ np.savez_compressed(out/'motion'/motion,paths=np.array(paths),t=np.array([f[0] for f in frames]),
160
+ positions=np.stack([f[1] for f in frames]),orientations=np.stack([f[2] for f in frames]))
161
+ report['motion']=f'motion/{motion}.npz'
162
+ return report
163
+
164
+ contacts=reset()
165
+ start_positions=array(view.get_world_poses()[0])
166
+ masses=array(view.get_masses())
167
+ direction,magnitude=context.get_gravity()
168
+ gravity=np.asarray(direction,dtype=float)*float(magnitude)
169
+ passive=rollout(3,contacts,'passive')
170
+ print('PHYSICS_STAGE passive',passive,flush=True)
171
+ end_positions=array(view.get_world_poses()[0])
172
+ moved=np.linalg.norm(end_positions-start_positions,axis=1)
173
+
174
+ manifest=json.loads((root/'objects.json').read_text())
175
+ manifest=manifest['objects'] if isinstance(manifest,dict) else manifest
176
+ result['manifest_errors']=[f"{o.get('id')}: label is missing or empty" for o in manifest if not str(o.get('label') or '').strip()]
177
+ result['manifest_errors']+=[f"{o.get('id')}: prim_path does not resolve" for o in manifest if not stage.GetPrimAtPath(o.get('prim_path','') or '/')]
178
+ def owned(obj):
179
+ prefix=str(obj.get('prim_path','')).rstrip('/')
180
+ return [i for i,p in enumerate(paths) if prefix and (p==prefix or p.startswith(prefix+'/'))]
181
+ drift=[{'id':o['id'],'dynamic_bodies':len(owned(o)),'center_drift_m':float(moved[owned(o)].max()),'mass_kg':float(masses[owned(o)].sum())}
182
+ for o in manifest if owned(o)]
183
+
184
+ requested=set(request['object_ids'])
185
+ eligible=[o for o in manifest if o['id'] in requested]
186
+ drops=[]
187
+ for obj in eligible:
188
+ free=[i for i in owned(obj) if paths[i] not in jointed]
189
+ if not free:continue
190
+ contacts=reset()
191
+ positions,orientations=view.get_world_poses()
192
+ raised=positions.clone() if hasattr(positions,'clone') else positions.copy()
193
+ raised[free,2]+=.05
194
+ view.set_world_poses(positions=raised,orientations=orientations)
195
+ applied=float((array(view.get_world_poses()[0])[free,2]-array(positions)[free,2]).min())
196
+ drops.append({'id':obj['id'],'input':'raise free bodies 0.05 m and release','bodies':[paths[i] for i in free],
197
+ 'lift_applied_m':applied,**rollout(2,contacts,f'drop_{len(drops)}')})
198
+ print('PHYSICS_STAGE drop',obj['id'],applied,drops[-1]['peak_linear_speed_mps'],flush=True)
199
+
200
+ # The common 2 ms step last, so the rollouts above run at the authored step.
201
+ result['realtime_common_step']={**measure_rtf(.002),'scope':'3 s passive rollout from the authored state at the common 2 ms step'}
202
+ print('PHYSICS_STAGE realtime_common_step',result['realtime_common_step']['rtf'],flush=True)
203
+ gravity_ok=bool(np.allclose(gravity,[0,0,-9.81],atol=.1))
204
+ mass_ok=bool(np.isfinite(masses).all() and (masses>=0).all())
205
+ result.update(status='completed',passive=passive,gravity_m_s2=gravity.tolist(),gravity_valid=gravity_ok,
206
+ mass_valid=mass_ok,center_drift=drift,drop_tests=drops,eligible_native_objects=len(eligible),
207
+ largest_body_displacements=[{'prim':paths[i],'displacement_m':float(moved[i])} for i in np.argsort(-moved)[:5]],
208
+ valid=bool(passive['valid'] and gravity_ok and mass_ok),
209
+ drop_scope='the requested objects with free (unjointed) dynamic bodies, in manifest order',
210
+ contact_scope=('contacts between each dynamic body and every collider (static, kinematic and dynamic; colliders '
211
+ 'inside a rigid body are filtered by that body, as GPU contact views require); '
212
+ 'static-static pairs are not simulated, as in MuJoCo world-body geoms'),
213
+ limitations=['the limits are the MuJoCo ones, not calibrated for PhysX',
214
+ 'kinematic, disabled and nested rigid bodies are not traced (see excluded_bodies)',
215
+ 'drift is per dynamic body under the object prim, not the object bounding box',
216
+ 'contacts with deformable bodies are not measured (GPU contact views cannot filter them)'])
217
+ except Exception as error:
218
+ result['error']=f'{type(error).__name__}: {error}'
219
+ finally:
220
+ (out/'physics.json').write_text(json.dumps(result,indent=2)+'\n')
221
+ if 'passive' in result:
222
+ passive=result['passive']
223
+ print('PASSIVE CHECK','PASSED' if passive['valid'] and not result.get('manifest_errors') else 'FAILED',
224
+ f"- {len(passive.get('initial_pairs',[]))} contact pairs > 5 mm, {passive['penetration_time_s']:.3f} s above 5 mm,"
225
+ f" peak speed {passive['peak_linear_speed_mps']:.2f} m/s, {len(result.get('manifest_errors',[]))} manifest errors;"
226
+ f" details in {out/'physics.json'}",flush=True)
227
+ else:print('CHECK FAILED:',result.get('error'),flush=True)
228
+ app.close()
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/input/video.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1cf7b26f24b4da06fdcc096e9f35f821a0733c1bfc71ad76f8fcb5805b9b22f3
3
+ size 11100373
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/build_scene.py ADDED
@@ -0,0 +1,551 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Reconstruct the meeting room in video.mp4 as a sim-ready Isaac Sim USD stage.
2
+
3
+ Run: /isaac-sim/python.sh /work/output/build_scene.py
4
+ Produces: scene.usda, textures/*.png, objects.json (self-contained, Z-up, metres).
5
+ """
6
+ from isaacsim import SimulationApp
7
+ app = SimulationApp({"headless": True, "multi_gpu": False, "active_gpu": 0,
8
+ "physics_gpu": 0, "enable_crashreporter": False,
9
+ "limit_cpu_threads": 8})
10
+
11
+ import json, math, os, random
12
+ from pathlib import Path
13
+ from PIL import Image, ImageDraw
14
+ from pxr import (Usd, UsdGeom, UsdShade, UsdPhysics, UsdLux, Sdf, Gf, Vt,
15
+ PhysxSchema)
16
+
17
+ OUT = Path("/work/output")
18
+ TEX = OUT / "textures"
19
+ TEX.mkdir(parents=True, exist_ok=True)
20
+ random.seed(7)
21
+
22
+ # ----------------------------------------------------------------------------- metrics
23
+ W, D, H = 4.60, 4.00, 2.75 # room interior X, Y, Z
24
+ NAVY = (0.094, 0.192, 0.302)
25
+ CREAM = (0.910, 0.886, 0.827)
26
+ CHAR = (0.188, 0.184, 0.173)
27
+ YELLOW = (0.945, 0.760, 0.090)
28
+ SILVER = (0.620, 0.650, 0.680)
29
+ OAK = (0.700, 0.545, 0.340)
30
+ EDGE = (0.170, 0.150, 0.130)
31
+ BIN = (0.895, 0.886, 0.845)
32
+ WHITE = (0.960, 0.960, 0.945)
33
+ BLIND = (0.720, 0.760, 0.790)
34
+ BLACK = (0.075, 0.080, 0.085)
35
+
36
+ stage = Usd.Stage.CreateNew(str(OUT / "scene.usda"))
37
+ UsdGeom.SetStageUpAxis(stage, UsdGeom.Tokens.z)
38
+ UsdGeom.SetStageMetersPerUnit(stage, 1.0)
39
+ UsdGeom.Xform.Define(stage, "/World")
40
+ stage.SetDefaultPrim(stage.GetPrimAtPath("/World"))
41
+
42
+ # ----------------------------------------------------------------------------- textures
43
+ def carpet_png():
44
+ im = Image.new("RGB", (512, 512), (104, 99, 92))
45
+ px = im.load()
46
+ for y in range(512):
47
+ for x in range(512):
48
+ t = 512 // 8
49
+ tx, ty = x // t, y // t
50
+ n = random.randint(-14, 14)
51
+ if (tx + ty) % 2 == 0:
52
+ s = (y % 3 == 0) * 26
53
+ else:
54
+ s = (x % 3 == 0) * 26
55
+ px[x, y] = (max(0, 104 + n - s), max(0, 99 + n - s), max(0, 92 + n - s))
56
+ im.save(TEX / "carpet.png")
57
+
58
+ def wood_png():
59
+ im = Image.new("RGB", (512, 256), (183, 143, 96))
60
+ d = ImageDraw.Draw(im)
61
+ for y in range(256):
62
+ c = random.randint(-16, 16)
63
+ d.line([(0, y), (512, y)], fill=(183 + c, 143 + c // 2, 96 + c // 3))
64
+ for _ in range(90):
65
+ y = random.randint(0, 255)
66
+ x0 = random.randint(0, 400)
67
+ d.line([(x0, y), (x0 + random.randint(30, 160), y + random.randint(-2, 2))],
68
+ fill=(150 + random.randint(-15, 15), 112, 66), width=1)
69
+ im.save(TEX / "wood.png")
70
+
71
+ def dots_png():
72
+ im = Image.new("RGB", (256, 256), (242, 198, 32))
73
+ d = ImageDraw.Draw(im)
74
+ for j in range(12):
75
+ for i in range(12):
76
+ x, y = 10 + i * 21 + (10 if j % 2 else 0), 10 + j * 21
77
+ d.ellipse([x, y, x + 7, y + 7], fill=(88, 80, 40))
78
+ im.save(TEX / "dots.png")
79
+
80
+ carpet_png(); wood_png(); dots_png()
81
+
82
+ # ----------------------------------------------------------------------------- materials
83
+ def mat_color(path, rgb, rough=0.55, metal=0.0, emissive=0.0, opacity=1.0):
84
+ m = UsdShade.Material.Define(stage, path)
85
+ s = UsdShade.Shader.Define(stage, path + "/Shader")
86
+ s.CreateIdAttr("UsdPreviewSurface")
87
+ s.CreateInput("diffuseColor", Sdf.ValueTypeNames.Color3f).Set(Gf.Vec3f(*rgb))
88
+ s.CreateInput("roughness", Sdf.ValueTypeNames.Float).Set(rough)
89
+ s.CreateInput("metallic", Sdf.ValueTypeNames.Float).Set(metal)
90
+ if emissive:
91
+ s.CreateInput("emissiveColor", Sdf.ValueTypeNames.Color3f).Set(
92
+ Gf.Vec3f(*(v * emissive for v in rgb)))
93
+ if opacity < 1.0:
94
+ s.CreateInput("opacity", Sdf.ValueTypeNames.Float).Set(opacity)
95
+ m.CreateSurfaceOutput().ConnectToSource(s.ConnectableAPI(), "surface")
96
+ return m
97
+
98
+ def mat_tex(path, png, rough=0.8):
99
+ m = UsdShade.Material.Define(stage, path)
100
+ s = UsdShade.Shader.Define(stage, path + "/Shader")
101
+ s.CreateIdAttr("UsdPreviewSurface")
102
+ s.CreateInput("roughness", Sdf.ValueTypeNames.Float).Set(rough)
103
+ t = UsdShade.Shader.Define(stage, path + "/Tex")
104
+ t.CreateIdAttr("UsdUVTexture")
105
+ t.CreateInput("file", Sdf.ValueTypeNames.Asset).Set("textures/" + png)
106
+ t.CreateInput("wrapS", Sdf.ValueTypeNames.Token).Set("repeat")
107
+ t.CreateInput("wrapT", Sdf.ValueTypeNames.Token).Set("repeat")
108
+ p = UsdShade.Shader.Define(stage, path + "/PVR")
109
+ p.CreateIdAttr("UsdPrimvarReader_float2")
110
+ p.CreateInput("varname", Sdf.ValueTypeNames.String).Set("st")
111
+ t.CreateInput("st", Sdf.ValueTypeNames.Float2).ConnectToSource(p.ConnectableAPI(), "result")
112
+ s.CreateInput("diffuseColor", Sdf.ValueTypeNames.Color3f).ConnectToSource(
113
+ t.ConnectableAPI(), "rgb")
114
+ m.CreateSurfaceOutput().ConnectToSource(s.ConnectableAPI(), "surface")
115
+ return m
116
+
117
+ M = {
118
+ "cream": mat_color("/World/Mats/Cream", CREAM, 0.85),
119
+ "navy": mat_color("/World/Mats/Navy", NAVY, 0.80),
120
+ "char": mat_color("/World/Mats/Charcoal", CHAR, 0.80),
121
+ "yellow": mat_color("/World/Mats/Yellow", YELLOW, 0.42),
122
+ "silver": mat_color("/World/Mats/Silver", SILVER, 0.30, 0.75),
123
+ "oak": mat_color("/World/Mats/Oak", OAK, 0.60),
124
+ "edge": mat_color("/World/Mats/Edge", EDGE, 0.50),
125
+ "bin": mat_color("/World/Mats/Bin", BIN, 0.55),
126
+ "white": mat_color("/World/Mats/White", WHITE, 0.60),
127
+ "blind": mat_color("/World/Mats/Blind", BLIND, 0.65),
128
+ "black": mat_color("/World/Mats/Black", BLACK, 0.55),
129
+ "glass": mat_color("/World/Mats/Glass", (0.80, 0.86, 0.90), 0.10, 0.0, opacity=0.25),
130
+ "light": mat_color("/World/Mats/Light", (1.0, 0.99, 0.94), 0.30, emissive=1.0),
131
+ "red": mat_color("/World/Mats/Red", (0.75, 0.20, 0.14), 0.5),
132
+ "wood": mat_tex("/World/Mats/WoodTex", "wood.png", 0.55),
133
+ "carpet": mat_tex("/World/Mats/CarpetTex", "carpet.png", 0.92),
134
+ "dots": mat_tex("/World/Mats/DotsTex", "dots.png", 0.45),
135
+ }
136
+
137
+ # ----------------------------------------------------------------------------- geometry helpers
138
+ objs = []
139
+
140
+ def cube(path, center, size, mat, collision=True, dynamic=False, mass=1.0,
141
+ rot=(0, 0, 0), invisible=False, friction=0.8):
142
+ p = UsdGeom.Cube.Define(stage, path)
143
+ p.CreateSizeAttr(1.0)
144
+ x = UsdGeom.Xformable(p)
145
+ x.AddTranslateOp().Set(Gf.Vec3d(*center))
146
+ if any(rot):
147
+ x.AddRotateXYZOp().Set(Gf.Vec3f(*rot))
148
+ x.AddScaleOp().Set(Gf.Vec3f(*size))
149
+ if invisible:
150
+ UsdGeom.Imageable(p).CreateVisibilityAttr().Set(UsdGeom.Tokens.invisible)
151
+ if mat:
152
+ UsdShade.MaterialBindingAPI.Apply(p.GetPrim()).Bind(mat)
153
+ if collision:
154
+ UsdPhysics.CollisionAPI.Apply(p.GetPrim())
155
+ if dynamic:
156
+ UsdPhysics.RigidBodyAPI.Apply(p.GetPrim())
157
+ UsdPhysics.MassAPI.Apply(p.GetPrim()).CreateMassAttr(mass)
158
+ return p
159
+
160
+ def quad(path, pts, mat, uv=(1.0, 1.0)):
161
+ """Flat textured quad from 4 corner points (no collision)."""
162
+ m = UsdGeom.Mesh.Define(stage, path)
163
+ m.CreatePointsAttr([Gf.Vec3f(*p) for p in pts])
164
+ m.CreateFaceVertexCountsAttr([4])
165
+ m.CreateFaceVertexIndicesAttr([0, 1, 2, 3])
166
+ n = (Gf.Vec3f(pts[1]) - Gf.Vec3f(pts[0])).GetNormalized() ^ (
167
+ Gf.Vec3f(pts[3]) - Gf.Vec3f(pts[0])).GetNormalized()
168
+ m.CreateNormalsAttr([n, n, n, n])
169
+ pv = UsdGeom.PrimvarsAPI(m.GetPrim()).CreatePrimvar(
170
+ "st", Sdf.ValueTypeNames.TexCoord2fArray, UsdGeom.Tokens.vertex)
171
+ pv.Set(Vt.Vec2fArray([(0, 0), (uv[0], 0), (uv[0], uv[1]), (0, uv[1])]))
172
+ UsdShade.MaterialBindingAPI.Apply(m.GetPrim()).Bind(mat)
173
+ return m
174
+
175
+ def xform(path, parent=None, pos=(0, 0, 0)):
176
+ p = UsdGeom.Xform.Define(stage, path)
177
+ UsdGeom.Xformable(p).AddTranslateOp().Set(Gf.Vec3d(*pos))
178
+ return p
179
+
180
+ def entry(o):
181
+ objs.append(o)
182
+
183
+ # ----------------------------------------------------------------------------- physics scene
184
+ sc = UsdPhysics.Scene.Define(stage, "/World/PhysicsScene")
185
+ sc.CreateGravityDirectionAttr(Gf.Vec3f(0, 0, -1))
186
+ sc.CreateGravityMagnitudeAttr(9.81)
187
+ try:
188
+ PhysxSchema.PhysxSceneAPI.Apply(sc.GetPrim()).CreateTimeStepsPerSecondAttr(120)
189
+ except Exception:
190
+ pass
191
+
192
+ # ----------------------------------------------------------------------------- shell
193
+ cube("/World/Shell/FloorSlab", (W / 2, D / 2, -0.05), (W, D, 0.10), M["char"],
194
+ friction=1.0)
195
+ quad("/World/Shell/Carpet",
196
+ [(0, 0, 0.001), (W, 0, 0.001), (W, D, 0.001), (0, D, 0.001)],
197
+ M["carpet"], uv=(W / 0.5, D / 0.5))
198
+ entry({"id": "carpet_floor", "label": "wall-to-wall striated carpet floor",
199
+ "prim_path": "/World/Shell/FloorSlab", "joint_paths": [],
200
+ "interaction_points": {}})
201
+
202
+ cube("/World/Shell/Ceiling", (W / 2, D / 2, H + 0.04), (W, D, 0.08), M["cream"])
203
+ entry({"id": "ceiling", "label": "ceiling", "prim_path": "/World/Shell/Ceiling",
204
+ "joint_paths": [], "interaction_points": {}})
205
+
206
+ def wall(name, center, size):
207
+ cube("/World/Shell/" + name, center, size, M["char"])
208
+ return "/World/Shell/" + name
209
+
210
+ # Four walls, split into lower (navy/charcoal) and upper (cream) bands.
211
+ wall("WestLow", (-0.04, D / 2, 0.52), (0.08, D, 1.04))
212
+ wall("WestUp", (-0.04, D / 2, 1.90), (0.08, D, 1.70))
213
+ wall("EastLow", (W + 0.04, D / 2, 0.52), (0.08, D, 1.04))
214
+ wall("EastUp", (W + 0.04, D / 2, 1.90), (0.08, D, 1.70))
215
+ wall("SouthLow", (W / 2, -0.04, 0.52), (W, 0.08, 1.04))
216
+ wall("SouthUp", (W / 2, -0.04, 1.90), (W, 0.08, 1.70))
217
+ wall("NorthLow", (W / 2, D + 0.04, 0.52), (W, 0.08, 1.04))
218
+ wall("NorthUp", (W / 2, D + 0.04, 1.90), (W, 0.08, 1.70))
219
+ # Recolour upper cream, lower navy/charcoal by material rebind.
220
+ for nm in ["WestUp", "EastUp", "SouthUp", "NorthUp"]:
221
+ UsdShade.MaterialBindingAPI.Apply(stage.GetPrimAtPath("/World/Shell/" + nm)).Bind(M["cream"])
222
+ for nm in ["WestLow", "SouthLow", "NorthLow"]:
223
+ UsdShade.MaterialBindingAPI.Apply(stage.GetPrimAtPath("/World/Shell/" + nm)).Bind(M["navy"])
224
+ UsdShade.MaterialBindingAPI.Apply(stage.GetPrimAtPath("/World/Shell/EastLow")).Bind(M["char"])
225
+
226
+ # Charcoal accent panels + baseboards.
227
+ for nm, c, s in [
228
+ ("AccentW", (0.012, 2.90, 1.55), (0.024, 1.30, 1.10)),
229
+ ("AccentE", (W - 0.012, 1.10, 1.55), (0.024, 1.50, 1.10)),
230
+ ("AccentS", (1.10, 0.012, 1.60), (1.30, 0.024, 1.00)),
231
+ ("NavyStrip", (W - 0.012, 3.30, 1.55), (0.024, 0.30, 1.10)),
232
+ ]:
233
+ cube("/World/Shell/" + nm, c, s, M["navy"] if nm == "NavyStrip" else M["char"])
234
+ for nm, c, s in [("BB_W", (0.012, D / 2, 0.045), (0.024, D, 0.09)),
235
+ ("BB_E", (W - 0.012, D / 2, 0.045), (0.024, D, 0.09)),
236
+ ("BB_S", (W / 2, 0.012, 0.045), (W, 0.024, 0.09)),
237
+ ("BB_N", (W / 2, D - 0.012, 0.045), (W, 0.024, 0.09))]:
238
+ cube("/World/Shell/" + nm, c, s, M["black"])
239
+ entry({"id": "wall_shell", "label": "cream navy and charcoal color-blocked walls",
240
+ "prim_path": "/World/Shell", "joint_paths": [], "interaction_points": {}})
241
+
242
+ # Ceiling light panels.
243
+ for i, (cx, cy) in enumerate([(1.2, 1.3), (3.4, 1.3), (1.2, 3.0), (3.4, 3.0)]):
244
+ cube("/World/Shell/Light%02d" % i, (cx, cy, H - 0.01), (1.10, 0.55, 0.02), M["light"])
245
+ entry({"id": "ceiling_light_%d" % i, "label": "recessed ceiling light panel",
246
+ "prim_path": "/World/Shell/Light%02d" % i, "joint_paths": [],
247
+ "interaction_points": {}})
248
+
249
+ # ----------------------------------------------------------------------------- window + blinds (north wall)
250
+ WX0, WX1, WZ0, WZ1 = 1.15, 3.45, 1.18, 2.36
251
+ # surround: rebuild north wall in four pieces around opening
252
+ for child in ["NorthLow", "NorthUp"]:
253
+ stage.GetPrimAtPath("/World/Shell/" + child).SetActive(False)
254
+ cube("/World/Window/Sill", ((WX0 + WX1) / 2, D - 0.02, WZ0 - 0.045),
255
+ (WX1 - WX0 + 0.12, 0.10, 0.09), M["cream"])
256
+ cube("/World/Window/Top", ((WX0 + WX1) / 2, D - 0.02, WZ1 + 0.05),
257
+ (WX1 - WX0 + 0.12, 0.10, 0.10), M["cream"])
258
+ cube("/World/Window/Left", (WX0 - 0.05, D - 0.02, (WZ0 + WZ1) / 2),
259
+ (0.10, 0.10, WZ1 - WZ0), M["cream"])
260
+ cube("/World/Window/Right", (WX1 + 0.05, D - 0.02, (WZ0 + WZ1) / 2),
261
+ (0.10, 0.10, WZ1 - WZ0), M["cream"])
262
+ cube("/World/Window/BottomWall", (W / 2, D + 0.04, WZ0 / 2), (W, 0.08, WZ0), M["navy"])
263
+ cube("/World/Window/TopWall", (W / 2, D + 0.04, (WZ1 + H) / 2), (W, 0.08, H - WZ1), M["cream"])
264
+ quad("/World/Window/Glass",
265
+ [(WX0, D - 0.015, WZ0), (WX1, D - 0.015, WZ0),
266
+ (WX1, D - 0.015, WZ1), (WX0, D - 0.015, WZ1)], M["glass"])
267
+ NS = 26
268
+ for i in range(NS):
269
+ z = WZ0 + 0.02 + (WZ1 - WZ0 - 0.04) * i / (NS - 1)
270
+ cube("/World/Window/Slat%02d" % i, ((WX0 + WX1) / 2, D - 0.028, z),
271
+ (WX1 - WX0 - 0.02, 0.018, 0.026), M["blind"], collision=False)
272
+ for ci, xx in enumerate((WX0 + 0.05, WX1 - 0.05)):
273
+ cube("/World/Window/Cord%d" % ci, (xx, D - 0.035, (WZ0 + WZ1) / 2),
274
+ (0.006, 0.006, WZ1 - WZ0), M["white"], collision=False)
275
+ entry({"id": "window_blinds", "label": "window with closed horizontal venetian blinds",
276
+ "prim_path": "/World/Window", "joint_paths": [], "interaction_points": {}})
277
+
278
+ # ----------------------------------------------------------------------------- black slotted wall rail
279
+ def rail(name, center, length, vertical=False):
280
+ cube("/World/" + name + "/Bar", center,
281
+ (0.045, length, 0.075) if vertical else (length, 0.045, 0.075), M["black"])
282
+ n = max(6, int(length / 0.085))
283
+ for i in range(n):
284
+ off = -length / 2 + (i + 0.5) * length / n
285
+ c = (center[0] + (off if not vertical else 0),
286
+ center[1] + (0 if not vertical else off), center[2])
287
+ s = (0.030, 0.038, 0.045) if not vertical else (0.038, 0.030, 0.045)
288
+ cube("/World/%s/Slot%02d" % (name, i),
289
+ (c[0] - (0.024 if not vertical else 0), c[1], c[2]), s,
290
+ M["char"], collision=False)
291
+
292
+ rail("RailNorth", (1.30, D - 0.045, 0.95), 2.20)
293
+ rail("RailWest", (0.045, 1.25, 0.95), 1.70, vertical=True)
294
+ entry({"id": "slotted_wall_rail", "label": "black horizontal slotted wall rail",
295
+ "prim_path": "/World/RailNorth", "joint_paths": [], "interaction_points": {}})
296
+
297
+ # Wall plates / outlets.
298
+ for i, (c, red) in enumerate([((2.70, D - 0.045, 1.42), False),
299
+ ((3.60, D - 0.045, 1.42), False),
300
+ ((0.045, 2.10, 1.45), False),
301
+ ((W - 0.045, 2.60, 1.45), True)]):
302
+ cube("/World/Plates/P%02d" % i, c, (0.10, 0.018, 0.065), M["white"], collision=False)
303
+ cube("/World/Plates/C%02d" % i,
304
+ (c[0] - (0.018 if i < 3 else 0), c[1] - (0.012 if i < 3 else 0), c[2]),
305
+ (0.030, 0.019, 0.030), M["red"] if red else M["char"], collision=False)
306
+
307
+ # Narrow glazed entrance sidelight (west wall near corner).
308
+ cube("/World/Sidelight/FrameL", (0.02, 0.62, 1.35), (0.045, 0.03, 2.20), M["white"])
309
+ cube("/World/Sidelight/FrameR", (0.02, 0.98, 1.35), (0.045, 0.03, 2.20), M["white"])
310
+ quad("/World/Sidelight/Glass",
311
+ [(0.02, 0.64, 0.30), (0.02, 0.96, 0.30), (0.02, 0.96, 2.45), (0.02, 0.64, 2.45)],
312
+ M["glass"])
313
+ entry({"id": "entrance_sidelight", "label": "narrow glazed entrance sidelight",
314
+ "prim_path": "/World/Sidelight", "joint_paths": [], "interaction_points": {}})
315
+
316
+ # ----------------------------------------------------------------------------- table (dynamic rigid body)
317
+ TC = (2.30, 2.00)
318
+ UsdGeom.Xform.Define(stage, "/World/Table")
319
+ cube("/World/Table/Top", (TC[0], TC[1], 0.740), (2.20, 1.10, 0.045), M["oak"])
320
+ quad("/World/Table/Grain",
321
+ [(TC[0] - 1.09, TC[1] - 0.54, 0.764), (TC[0] + 1.09, TC[1] - 0.54, 0.764),
322
+ (TC[0] + 1.09, TC[1] + 0.54, 0.764), (TC[0] - 1.09, TC[1] + 0.54, 0.764)],
323
+ M["wood"], uv=(3.0, 1.0))
324
+ cube("/World/Table/Edge", (TC[0], TC[1], 0.712), (2.23, 1.13, 0.018), M["edge"],
325
+ collision=False)
326
+ cube("/World/Table/Pedestal", (TC[0], TC[1], 0.385), (1.55, 0.10, 0.665), M["char"])
327
+ cube("/World/Table/PedTrim", (TC[0], TC[1] - 0.056, 0.385), (1.55, 0.016, 0.665),
328
+ M["silver"], collision=False)
329
+ cube("/World/Table/FootLong", (TC[0], TC[1], 0.030), (1.95, 0.13, 0.055), M["silver"])
330
+ cube("/World/Table/FootCross", (TC[0], TC[1], 0.030), (0.13, 0.62, 0.055), M["silver"])
331
+ trb = stage.GetPrimAtPath("/World/Table")
332
+ UsdPhysics.RigidBodyAPI.Apply(trb)
333
+ UsdPhysics.MassAPI.Apply(trb).CreateMassAttr(28.0)
334
+ entry({"id": "meeting_table", "label": "large rectangular oak meeting table with metal T-base",
335
+ "prim_path": "/World/Table", "joint_paths": [],
336
+ "interaction_points": {"grasp": "/World/Table/Grasp"}})
337
+ xform("/World/Table/Grasp", pos=(TC[0] + 0.95, TC[1], 0.77))
338
+
339
+ # ----------------------------------------------------------------------------- chairs (6, dynamic)
340
+ CHAIRS = [(1.55, 1.02, 0), (3.05, 1.02, 0), (1.55, 2.98, 180),
341
+ (3.05, 2.98, 180), (0.80, 2.00, 90), (3.70, 2.00, -90)]
342
+
343
+ def chair(idx, cx, cy, yaw):
344
+ root = "/World/Chair%d" % idx
345
+ UsdGeom.Xform.Define(stage, root)
346
+ rad = math.radians(yaw)
347
+ co, si = math.cos(rad), math.sin(rad)
348
+ def P(lx, ly, lz):
349
+ return (cx + lx * co - ly * si, cy + lx * si + ly * co, lz)
350
+ def part(name, lx, ly, lz, sx, sy, sz, mat, coll=True, rot=0):
351
+ cube(root + "/" + name, P(lx, ly, lz), (sx, sy, sz), mat,
352
+ collision=coll, rot=(0, 0, yaw + rot))
353
+ # seat, back, legs, arms, crossbars
354
+ part("Seat", 0, 0, 0.452, 0.50, 0.46, 0.055, M["yellow"])
355
+ part("Back", 0, -0.215, 0.730, 0.46, 0.042, 0.52, M["yellow"], rot=-8)
356
+ # perforated texture on the back (visual only quad)
357
+ bx, by, bz = P(0, -0.238, 0.730)
358
+ q = [(bx - 0.115 * co, by - 0.115 * si, bz - 0.255),
359
+ (bx + 0.115 * co, by + 0.115 * si, bz - 0.255),
360
+ (bx + 0.115 * co, by + 0.115 * si, bz + 0.255),
361
+ (bx - 0.115 * co, by - 0.115 * si, bz + 0.255)]
362
+ quad(root + "/Perf", q, M["dots"], uv=(2.0, 2.0))
363
+ for sgn_x in (-1, 1):
364
+ for sgn_y in (-1, 1):
365
+ part("Leg%d%d" % (sgn_x + 1, sgn_y + 1), sgn_x * 0.205, sgn_y * 0.190,
366
+ 0.217, 0.034, 0.034, 0.428, M["silver"])
367
+ for sgn in (-1, 1):
368
+ part("Arm%d" % (sgn + 1), sgn * 0.255, 0.0, 0.620, 0.030, 0.40, 0.030,
369
+ M["silver"], coll=False)
370
+ part("BarFront", 0, 0.190, 0.090, 0.42, 0.028, 0.028, M["silver"], coll=False)
371
+ part("BarBack", 0, -0.190, 0.090, 0.42, 0.028, 0.028, M["silver"], coll=False)
372
+ # whole chair is one dynamic rigid body; every part is a child collider
373
+ rb = stage.GetPrimAtPath(root)
374
+ UsdPhysics.RigidBodyAPI.Apply(rb)
375
+ UsdPhysics.MassAPI.Apply(rb).CreateMassAttr(5.0)
376
+ xform(root + "/Grasp", pos=P(0, 0, 0.50))
377
+ entry({"id": "chair_%d" % idx, "label": "yellow perforated molded plastic caster chair",
378
+ "prim_path": root, "joint_paths": [],
379
+ "interaction_points": {"grasp": root + "/Grasp"}})
380
+
381
+ for i, (cx, cy, yaw) in enumerate(CHAIRS, 1):
382
+ chair(i, cx, cy, yaw)
383
+
384
+ # ----------------------------------------------------------------------------- built-in cabinet, 2 hinged doors
385
+ CX = 4.35
386
+ # hollow carcass (no front panel) — static
387
+ for nm, c, s in [
388
+ ("Back", (4.525, 1.50, 1.10), (0.05, 1.82, 2.10)),
389
+ ("Top", (CX, 1.50, 2.12), (0.36, 1.82, 0.05)),
390
+ ("Bottom", (CX, 1.50, 0.08), (0.36, 1.82, 0.05)),
391
+ ("SideL", (CX, 0.615, 1.10), (0.36, 0.05, 2.05)),
392
+ ("SideR", (CX, 2.385, 1.10), (0.36, 0.05, 2.05)),
393
+ ("Divider", (CX, 1.50, 1.10), (0.36, 0.04, 2.05)),
394
+ ("Shelf", (CX, 1.50, 1.10), (0.36, 1.78, 0.03)),
395
+ ]:
396
+ cube("/World/Cabinet/" + nm, c, s, M["cream"])
397
+
398
+ doors = []
399
+ for i, (yc, hy, lo, hi, hname) in enumerate([
400
+ (1.06, 0.60, -100.0, 0.0, "Left"),
401
+ (1.94, 2.40, 0.0, 100.0, "Right")]):
402
+ dp = "/World/Cabinet/Door" + hname
403
+ cube(dp, (4.13, yc, 1.10), (0.030, 0.86, 2.00), M["cream"],
404
+ dynamic=True, mass=6.0)
405
+ cube(dp + "/Seam", (4.113, yc, 1.10), (0.006, 0.80, 1.94), M["cream"],
406
+ collision=False)
407
+ jp = "/World/Cabinet/Joint" + hname
408
+ j = UsdPhysics.RevoluteJoint.Define(stage, jp)
409
+ j.CreateBody1Rel().SetTargets([dp])
410
+ j.CreateLocalPos0Attr().Set(Gf.Vec3f(4.13, hy, 1.10))
411
+ j.CreateLocalPos1Attr().Set(Gf.Vec3f(0, yc - hy, 0))
412
+ j.CreateAxisAttr().Set("Z")
413
+ j.CreateLowerLimitAttr().Set(lo)
414
+ j.CreateUpperLimitAttr().Set(hi)
415
+ xform(dp + "/Handle", pos=(4.105, yc + (0.32 if i == 0 else -0.32), 1.05))
416
+ xform(dp + "/handle", pos=(4.105, yc + (0.32 if i == 0 else -0.32), 1.05))
417
+ doors.append({"id": "cabinet_door_" + hname.lower(),
418
+ "label": "cream flat-panel hinged cabinet door",
419
+ "prim_path": dp, "joint_paths": [jp],
420
+ "interaction_points": {"handle": dp + "/Handle"}})
421
+ entry(doors[-1])
422
+ entry({"id": "built_in_cabinet", "label": "cream flat-panel built-in cabinet carcass",
423
+ "prim_path": "/World/Cabinet", "joint_paths": [], "interaction_points": {}})
424
+
425
+ # ----------------------------------------------------------------------------- low credenza with a drawer
426
+ CCY = 3.15
427
+ for nm, c, s in [
428
+ ("Back", (0.105, CCY, 0.46), (0.05, 1.00, 0.80)),
429
+ ("Top", (0.29, CCY, 0.83), (0.36, 1.00, 0.05)),
430
+ ("Bottom", (0.29, CCY, 0.09), (0.36, 1.00, 0.05)),
431
+ ("SideL", (0.29, CCY - 0.49, 0.46), (0.36, 0.05, 0.80)),
432
+ ("SideR", (0.29, CCY + 0.49, 0.46), (0.36, 0.05, 0.80)),
433
+ ]:
434
+ cube("/World/Credenza/" + nm, c, s, M["cream"])
435
+ dr = "/World/Credenza/Drawer"
436
+ cube(dr, (0.30, CCY, 0.50), (0.30, 0.88, 0.28), M["cream"], dynamic=True, mass=4.0)
437
+ cube(dr + "/Front", (0.452, CCY, 0.50), (0.020, 0.90, 0.34), M["cream"],
438
+ collision=False)
439
+ jp = "/World/Credenza/DrawerJoint"
440
+ j = UsdPhysics.PrismaticJoint.Define(stage, jp)
441
+ j.CreateBody1Rel().SetTargets([dr])
442
+ j.CreateLocalPos0Attr().Set(Gf.Vec3f(0.30, CCY, 0.50))
443
+ j.CreateLocalPos1Attr().Set(Gf.Vec3f(0, 0, 0))
444
+ j.CreateAxisAttr().Set("X")
445
+ j.CreateLowerLimitAttr().Set(0.0)
446
+ j.CreateUpperLimitAttr().Set(0.30)
447
+ xform(dr + "/Handle", pos=(0.465, CCY, 0.50))
448
+ entry({"id": "credenza_drawer", "label": "sliding drawer of low cream credenza",
449
+ "prim_path": dr, "joint_paths": [jp],
450
+ "interaction_points": {"handle": dr + "/Handle"}})
451
+ entry({"id": "credenza", "label": "low cream credenza carcass",
452
+ "prim_path": "/World/Credenza", "joint_paths": [], "interaction_points": {}})
453
+
454
+ # ----------------------------------------------------------------------------- scale anchor: trash bin (hollow)
455
+ BX, BY = 0.55, 0.55
456
+ BL, BW, BH = 0.418, 0.294, 0.483 # exact anchored extents
457
+ T = 0.014
458
+ broot = "/World/TrashBin"
459
+ UsdGeom.Xform.Define(stage, broot)
460
+ UsdGeom.Xformable(stage.GetPrimAtPath(broot)).AddTranslateOp().Set(Gf.Vec3d(0, 0, 0.002))
461
+ for nm, c, s in [
462
+ ("WallN", (BX, BY + BW / 2 - T / 2, BH / 2), (BL, T, BH)),
463
+ ("WallS", (BX, BY - BW / 2 + T / 2, BH / 2), (BL, T, BH)),
464
+ ("WallE", (BX + BL / 2 - T / 2, BY, BH / 2), (T, BW - 2 * T, BH)),
465
+ ("WallW", (BX - BL / 2 + T / 2, BY, BH / 2), (T, BW - 2 * T, BH)),
466
+ ("Bottom", (BX, BY, T / 2), (BL - 2 * T, BW - 2 * T, T)),
467
+ ]:
468
+ cube(broot + "/" + nm, c, s, M["bin"])
469
+ # tapered rim lip
470
+ cube(broot + "/Rim", (BX, BY, BH - 0.004), (BL + 0.010, BW + 0.010, 0.016),
471
+ M["white"], collision=False)
472
+ rb = stage.GetPrimAtPath(broot)
473
+ UsdPhysics.RigidBodyAPI.Apply(rb)
474
+ UsdPhysics.MassAPI.Apply(rb).CreateMassAttr(1.6)
475
+ xform(broot + "/Grasp", pos=(BX, BY, BH * 0.6))
476
+ entry({"id": "scale_anchor", "label": "cream rectangular waste bin (scale anchor 0.418x0.294x0.483 m)",
477
+ "prim_path": broot, "joint_paths": [],
478
+ "interaction_points": {"grasp": broot + "/Grasp"}})
479
+
480
+ # ----------------------------------------------------------------------------- deformable: bin liner (native cloth)
481
+ def liner():
482
+ p = "/World/BinLiner"
483
+ n = 11
484
+ pts, uvs = [], []
485
+ for j in range(n):
486
+ for i in range(n):
487
+ u, v = i / (n - 1), j / (n - 1)
488
+ x = BX + (u - 0.5) * (BL + 0.03)
489
+ y = BY + (v - 0.5) * (BW + 0.03)
490
+ # drape: centre dips into the bin, edges hang over the rim
491
+ edge = min(u, 1 - u, v, 1 - v)
492
+ z = BH + 0.012 - 0.16 * (edge * 4) ** 0.6
493
+ pts.append(Gf.Vec3f(x, y, z))
494
+ uvs.append((u * 2.0, v * 2.0))
495
+ idx = []
496
+ for j in range(n - 1):
497
+ for i in range(n - 1):
498
+ a = j * n + i
499
+ idx.extend([a, a + 1, a + n + 1, a, a + n + 1, a + n])
500
+ m = UsdGeom.Mesh.Define(stage, p)
501
+ m.CreatePointsAttr(pts)
502
+ m.CreateFaceVertexCountsAttr([3] * (len(idx) // 3))
503
+ m.CreateFaceVertexIndicesAttr(idx)
504
+ m.CreateNormalsAttr([Gf.Vec3f(0, 0, 1)] * len(pts))
505
+ pv = UsdGeom.PrimvarsAPI(m.GetPrim()).CreatePrimvar(
506
+ "st", Sdf.ValueTypeNames.TexCoord2fArray, UsdGeom.Tokens.vertex)
507
+ pv.Set(Vt.Vec2fArray(uvs))
508
+ UsdShade.MaterialBindingAPI.Apply(m.GetPrim()).Bind(M["white"])
509
+ # native Omni/PhysX deformable surface schemas (generic applied-schema authoring)
510
+ prim = m.GetPrim()
511
+ for sch in ("OmniPhysicsBodyAPI", "OmniPhysicsDeformableBodyAPI",
512
+ "OmniPhysicsSurfaceDeformableSimAPI", "PhysxAutoDeformableBodyAPI"):
513
+ try:
514
+ prim.AddAppliedSchema(sch)
515
+ except Exception:
516
+ pass
517
+ # deformable surface material
518
+ try:
519
+ mp = "/World/Mats/LinerDeformable"
520
+ mm = UsdShade.Material.Define(stage, mp)
521
+ mm.GetPrim().AddAppliedSchema("OmniPhysicsDeformableMaterialAPI")
522
+ mm.GetPrim().AddAppliedSchema("OmniPhysicsSurfaceDeformableMaterialAPI")
523
+ UsdShade.MaterialBindingAPI.Apply(prim).Bind(mm)
524
+ except Exception:
525
+ pass
526
+ xform("/World/BinLiner/Grasp", pos=(BX, BY, BH + 0.02))
527
+ entry({"id": "bin_liner", "label": "white soft plastic bin liner (deformable cloth)",
528
+ "prim_path": p, "joint_paths": [],
529
+ "interaction_points": {"grasp": "/World/BinLiner/Grasp"},
530
+ "material_corners": [0, n - 1, n * n - 1, n * (n - 1)]})
531
+
532
+ try:
533
+ liner()
534
+ except Exception as e:
535
+ print("LINER_WARNING", e)
536
+
537
+ # ----------------------------------------------------------------------------- lighting
538
+ dome = UsdLux.DomeLight.Define(stage, "/World/Dome")
539
+ dome.CreateIntensityAttr(700.0)
540
+ for i, (cx, cy) in enumerate([(1.3, 1.6), (3.2, 2.6)]):
541
+ rl = UsdLux.RectLight.Define(stage, "/World/Key%d" % i)
542
+ rl.CreateIntensityAttr(900.0)
543
+ rl.CreateWidthAttr(1.4); rl.CreateHeightAttr(0.9)
544
+ UsdGeom.Xformable(rl).AddTranslateOp().Set(Gf.Vec3d(cx, cy, H - 0.12))
545
+ UsdGeom.Xformable(rl).AddRotateXYZOp().Set(Gf.Vec3f(0, 180, 0))
546
+
547
+ # ----------------------------------------------------------------------------- write
548
+ (OUT / "objects.json").write_text(json.dumps({"objects": objs}, indent=2) + "\n")
549
+ stage.GetRootLayer().Save()
550
+ print("BUILT", len(objs), "objects;", "scene.usda")
551
+ app.close()
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/scene.usda ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ #usda 1.0
2
+
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/carpet.png ADDED

Git LFS Details

  • SHA256: 7bfc9936ef9f7fff145ba62ea4874a9121e2d96439ee0f1e155ca183e1be570e
  • Pointer size: 131 Bytes
  • Size of remote file: 307 kB
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/dots.png ADDED

Git LFS Details

  • SHA256: 98141cc59dd914edbe08181948c11e97bc15a23c718f74e5be169f59fb1a80fa
  • Pointer size: 129 Bytes
  • Size of remote file: 1.03 kB
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/output/textures/wood.png ADDED

Git LFS Details

  • SHA256: 4910c81c03eb87eef98ff10fd661deb6468e3b0fbb4ca2dca331f0e965975846
  • Pointer size: 129 Bytes
  • Size of remote file: 4.6 kB
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/run.json ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "track": "scenes",
3
+ "engine": "isaac",
4
+ "item": "rcd_yellow_chair_meeting_room",
5
+ "run_id": "20261114T065536Z",
6
+ "campaign": "jiarui-2-codebuddy",
7
+ "cli": "codebuddy",
8
+ "model": "Hunyuan 4",
9
+ "model_id": "hy4-preview",
10
+ "model_family": "hy4-preview",
11
+ "model_reasoning_effort": "default",
12
+ "budget_seconds": 1800,
13
+ "told_budget_seconds": 1800,
14
+ "stage": 1,
15
+ "continued_from": null,
16
+ "contract": {
17
+ "name": "scenes-isaac",
18
+ "sha256": "f1ab7ea88842502d37176b41d9f3bec64a5d9fac292ca3f16caf08da3b5db96d",
19
+ "track": "scenes",
20
+ "engine": "isaac",
21
+ "engine_version": "6.2.0",
22
+ "image": "simready-isaac-generation:latest",
23
+ "blender_version": "5.2.3",
24
+ "inputs": [
25
+ "video.mp4",
26
+ "scale_anchor.json",
27
+ "anchor_reference.png",
28
+ "anchor_marked.png",
29
+ "prompt.txt",
30
+ "generation_spec.json",
31
+ "self_check.py",
32
+ "sdk_smoke.py"
33
+ ],
34
+ "withheld": [
35
+ "groundtruth/ (objects, cameras, task catalogs, object dynamics, interaction test) and its source tag",
36
+ "provenance/ (source annotation, anchor selection)",
37
+ "the frozen protocol of the track and engine (data/benchmark/frozen/<track>/<engine>.json)",
38
+ "other runs"
39
+ ],
40
+ "outputs": [
41
+ "scene.usda",
42
+ "build_scene.py",
43
+ "objects.json",
44
+ "render_preview.py",
45
+ "preview.mp4",
46
+ "interaction.mp4",
47
+ "README.md",
48
+ "refinement_log.json"
49
+ ],
50
+ "self_check": "/isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check",
51
+ "budget": {
52
+ "told": "the first of the session's cumulative budgets, in wall-clock seconds from container start",
53
+ "resume": "each later budget resumes the same conversation for the rest of its time; each budget is its own run",
54
+ "enforcement": "the runner kills the container at the deadline; the output as it stands is the result"
55
+ },
56
+ "refinement": {
57
+ "minimum_completed_rounds": 3,
58
+ "minimum_reference_views": 6,
59
+ "minimum_inspected_video_frames": 12,
60
+ "final_validation_seconds": 90
61
+ },
62
+ "isolation": {
63
+ "container": "Docker: read-only input (and tools), writable output and scratch, temporary credentials; no repository, benchmark ground truth or evaluator files",
64
+ "network": "an internal network whose only exit is an HTTPS CONNECT proxy (runtime/egress_proxy.py)",
65
+ "egress_hosts": [
66
+ "api.anthropic.com",
67
+ "platform.claude.com",
68
+ "chatgpt.com",
69
+ "auth.openai.com",
70
+ "opencode.ai",
71
+ "mcp.exa.ai",
72
+ "search.parallel.ai"
73
+ ]
74
+ },
75
+ "protocol": {
76
+ "file": "data/benchmark/frozen/scenes/isaac.json",
77
+ "sha256": "df0fe5bb1c796474ad13cdba81b660a2946aa24297cb289f9581cfde7c035bb7"
78
+ }
79
+ },
80
+ "engine_version": "6.2.0",
81
+ "image": "simready-isaac-generation:latest",
82
+ "image_id": "sha256:8d9f9c934457aa3c8e91e8099e35db27b50db287d0eb76a647052c310e754d87",
83
+ "required_refinement_rounds": 3,
84
+ "gpu_uuid": "GPU-ab037b41-847d-64be-b2b8-ba682a201943",
85
+ "egress_mode": "internal_network",
86
+ "input_snapshot": "input",
87
+ "tools": null,
88
+ "evaluator_protocol": "evaluator_protocol.json",
89
+ "output": "output",
90
+ "log": "generation.jsonl",
91
+ "status": "budget_expired",
92
+ "agent_cli": {
93
+ "cli": "codebuddy",
94
+ "name": "CodeBuddy Code",
95
+ "binary": "a pinned codebuddy-code package and its node, mounted read-only (SIMREADY_CODEBUDDY_PREFIX, default the host's codebuddy install)",
96
+ "login": "the CodeBuddy login file, given to the CLI as a snapshot without its refresh token (the session cannot rotate it)",
97
+ "billing": "CodeBuddy account (the CLI reports no price)",
98
+ "tools": "every tool but WebFetch and the server-side media generators ImageGen, ImageEdit, VideoGen (--disallowedTools); WebSearch runs on CodeBuddy's side",
99
+ "hosts": [
100
+ "[redacted-host]",
101
+ "wb.[redacted]buddy.com"
102
+ ],
103
+ "version": "2.182.0",
104
+ "model": "hy4-preview",
105
+ "model_family": "hy4-preview"
106
+ },
107
+ "started_at": "2026-11-14T06:55:37.233413+00:00",
108
+ "model_answered": true,
109
+ "exit_code": 137,
110
+ "credential_refresh_persistence": "snapshot_without_refresh_token",
111
+ "finished_at": "2026-11-14T07:25:39.852835+00:00",
112
+ "cli_session_id": "01a1d8d7-2060-754d-a23b-34bf42dc6672",
113
+ "api_retries": 0,
114
+ "rate_limit_events": 0,
115
+ "stage_seconds": 1802.729,
116
+ "elapsed_seconds": 1802.729,
117
+ "cost": {
118
+ "status": "partial",
119
+ "source": "the CLI's running usage totals (usage_update events)",
120
+ "usd": null,
121
+ "usd_basis": "not reported (CodeBuddy account billing)",
122
+ "credits": 0,
123
+ "credit_status": "known",
124
+ "turns": null,
125
+ "api_seconds": null,
126
+ "tokens": {
127
+ "input": 115172,
128
+ "cache_read": 1343488,
129
+ "cache_write": 0,
130
+ "output": 48872,
131
+ "reasoning": null,
132
+ "total": 1507532
133
+ },
134
+ "wall_seconds": 1802.729,
135
+ "budget_seconds": 1800,
136
+ "stopped_at_budget": true
137
+ },
138
+ "scene_present": true
139
+ }
scenes/isaac/rcd_yellow_chair_meeting_room/20261114T065536Z/submitted_prompt.txt ADDED
@@ -0,0 +1,143 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ About the simulator. NVIDIA Isaac Sim is a robotics simulator built on Omniverse. You describe a scene
2
+ as an OpenUSD stage: prims carrying physics schemas for rigid bodies, colliders, joints and articulations, and
3
+ deformable bodies and surfaces. NVIDIA PhysX simulates it, here with GPU dynamics, and RTX path tracing renders it.
4
+ Isaac Sim is widely used for robot learning and sim-to-real work.
5
+
6
+ Reconstruct the room shown in video.mp4 as an interactive Isaac Sim scene.
7
+
8
+ THE RESULT MUST BE SIM-READY, READY FOR INTERACTIVE ROBOT LEARNING. It is an environment a robot learns
9
+ in, not a picture of a room. A training pipeline must be able to load it headless, reset it, step it for
10
+ a long time and let robots act on it, with no manual repair. That means:
11
+ - It loads with no network access or external files, and resets to the same state at rest.
12
+ - Everything a robot could touch has collision geometry matching what is visible. An object with no
13
+ collision does not exist for a robot, however good it looks.
14
+ - Rigid things are rigid bodies with plausible mass; free objects can be pushed, and doors, drawers and lids
15
+ move on joints with limits.
16
+ - Soft things are native deformables: cloth, bedding, curtains, rugs, throws, cushions and pillows.
17
+ A rigid body that merely looks soft is wrong.
18
+ - It stays stable under sustained interaction, across many steps without a reset, and steps fast
19
+ enough to train on.
20
+
21
+ Budget: 30 minutes (1,800 seconds) of wall-clock time per scene, starting when
22
+ you receive access to the input files. All inspection, coding, asset creation,
23
+ tool calls, rendering, debugging, and validation must fit within this budget.
24
+ Stop work at the deadline and submit the latest valid, loadable result.
25
+ Do not stop after producing the first runnable scene. Use the available time
26
+ for repeated self-critique and refinement; reserve the final 90 seconds for
27
+ validation and submission.
28
+
29
+ Inputs (in /work/input):
30
+ - video.mp4: the complete source video.
31
+ - scale_anchor.json: dimensions and identity of one object in the room, the scale anchor.
32
+ - anchor_reference.png and anchor_marked.png: images identifying that object.
33
+ The marked rectangle is a projected box envelope, not a segmentation mask.
34
+ - generation_spec.json: this contract in machine-readable form.
35
+ - self_check.py: the public self-check described below.
36
+
37
+ Maximize visual realism and fidelity to the source video. Reproduce the room's
38
+ geometry, proportions, native objects, materials, textures, colors, lighting,
39
+ and visible details as faithfully as possible within the time budget. Use the
40
+ scale anchor to establish metric scale; infer other dimensions and layout from
41
+ the video. The anchor's known real dimensions let you work out every other
42
+ object's size and the distances between objects, for example by measuring them
43
+ against the anchor in frames where both are visible, or by scaling a
44
+ reconstruction so that the anchor has its true size. Place the anchor object
45
+ where it is in the video; the room is not aligned to it alone.
46
+
47
+ Visual objective: MAXIMIZE THE VISUAL QUALITY of the reconstructed room, without giving up any
48
+ sim-ready requirement: a scene that looks right but cannot be simulated has failed.
49
+ Use source-aligned renders to improve what a viewer actually sees: recognizable
50
+ furniture silhouettes, correct proportions and placement, detailed surfaces,
51
+ plausible texture scale, window appearance, soft lighting and small distinctive
52
+ objects. Avoid a room made only of plain boxes when the video shows richer
53
+ shapes and materials. Create local textures or meshes when they improve fidelity;
54
+ do not use a single photograph or video plane as a substitute for 3D geometry.
55
+ Check the entire camera path, including corners and reverse views, rather than
56
+ optimizing one attractive render.
57
+
58
+ Required self-critique and refinement:
59
+ 1. Inspect at least 12 frames distributed across the video. Choose at least
60
+ six representative reference views covering the room and important objects.
61
+ Record their timestamps. Estimate corresponding cameras using ONLY the
62
+ public video and scale anchor, and reuse those cameras across iterations.
63
+ 2. Build an early loadable baseline and render those views. Do not spend the
64
+ whole budget on a first draft or a single attractive camera angle.
65
+ 3. Complete at least THREE render-critique-refine-verify rounds. In each round,
66
+ open side-by-side source/render contact sheets with the image-capable Read
67
+ tool. Identify concrete visible discrepancies, implement the highest-impact
68
+ fixes, then re-render the SAME views and check for regressions. Writing a
69
+ critique without changing and re-rendering the scene is not a completed round.
70
+ 4. Round 1 prioritizes room shape, scale, furniture placement, object dimensions
71
+ and silhouettes. Round 2 prioritizes materials, textures, lighting, colors,
72
+ windows and salient details. Round 3 revisits visual mismatches and verifies
73
+ collision support, movable parts and soft items, and runs the self-check.
74
+ 5. Continue additional rounds while time permits. Do not wait until the deadline
75
+ to save a valid scene. Report the true number of completed rounds.
76
+ 6. After each round, save a valid checkpoint and record the source timestamps,
77
+ render paths, specific discrepancies, actual edits, before/after observations,
78
+ and remaining issues in refinement_log.json. Treat self-assessments as
79
+ diagnostics.
80
+ 7. Keep the best validated complete scene, and restore it if a later change
81
+ regresses. The final output must describe that selected result.
82
+
83
+ You may use any tools, including web search, and the software installed here,
84
+ including Blender (`blender`). Download nothing from the internet (no files,
85
+ models, assets, datasets or packages), and do not look up this room, its video
86
+ or the dataset it comes from. Use no annotations, scans or other data about this
87
+ room beyond your inputs, no previous generated outputs and no other session's
88
+ reconstruction. Do not read authentication files. Do not add a robot.
89
+
90
+ The result must be a genuinely interactive 3D scene:
91
+ - Preserve native object identities, count, dimensions, and spatial relationships.
92
+ - Provide collision geometry consistent with visible surfaces and free spaces.
93
+ - Separate movable objects and articulated parts, with appropriate joints and limits.
94
+ - Build every visible soft item with the simulator's native deformable mechanism, as one object
95
+ entry each, with a label that says plainly what it is, for example "curtain".
96
+ - Preserve openings, container interiors, support surfaces, and access needed
97
+ for interactions. Do not replace the room's own objects with generic stand-ins.
98
+ - Assign physically plausible properties and document estimates for unknowns;
99
+ the video is not measured mass, friction or stiffness.
100
+
101
+ Construction rules (the self-check reports them):
102
+ - The scene must start at rest without interpenetration: no contact deeper than
103
+ 5 mm between any two bodies or between a body and fixed geometry.
104
+ - Articulated parts (drawers, doors, lids) must not intersect their own carcass in
105
+ any position of their range: hollow the carcass or exclude the pair from collision.
106
+ - Door and drawer swing paths must be clear of other furniture.
107
+ - The timestep is your choice.
108
+
109
+ Object labels: each objects.json entry has a "label" that says plainly what the object is
110
+ (for example "double bed", "floor lamp"). Use one entry per physical instance.
111
+
112
+ Interaction points: for articulated furniture give a "handle" point on each moving
113
+ part; for movable graspable objects a "grasp" point; for pourable containers a
114
+ "spout" point; for a visible bed cover, four "material_corners" in boundary order.
115
+
116
+ Self-check: run it on your output after each round and before submitting. It loads
117
+ only your output, simulates 3 s from the authored state and reports contacts deeper
118
+ than 5 mm (with the two bodies involved), the time above 5 mm, peak body speed,
119
+ non-finite state and entries without a label. The scene must stay within these
120
+ limits: at most 0.02 s above 5 mm, peak speed at most 5 m/s, finite state. Fix what it reports; do not edit it.
121
+ /isaac-sim/python.sh /work/input/self_check.py --scene /work/output --out /tmp/self_check
122
+
123
+ Output format (Isaac Sim, in /work/output):
124
+ - scene.usda (or scene.usd): self-contained USD stage, Z-up, metres, explicit default
125
+ prim, locally referenced assets and textures, and an authored PhysX scene.
126
+ - build_scene.py: reproducible builder using the installed Isaac Sim / pxr APIs.
127
+ - objects.json: {"objects": [...]}, one entry per physical object instance:
128
+ "id": a persistent unique string (keep the scale anchor ID),
129
+ "label": what the object is, in plain words,
130
+ "prim_path": the object's root prim (no single mesh containing all furniture),
131
+ "joint_paths": articulated joint prims, if present,
132
+ "interaction_points": interaction Xform prim paths, named handle, grasp or spout
133
+ (child Xforms of the moving part or object),
134
+ "material_corners": four mesh vertex indices for a deformable bed cover.
135
+ - render_preview.py, at least six views rendered with RTX PathTracing, preview.mp4
136
+ (camera tour) and interaction.mp4 (native physics interaction with an original object).
137
+ - README.md and refinement_log.json, as above.
138
+ Runtime: /isaac-sim/python.sh YOUR_SCRIPT.py runs the bundled Python with Isaac Sim,
139
+ USD, NumPy and Pillow; /usr/local/bin/ffmpeg and blender are available. Use headless=True,
140
+ multi_gpu=False, active_gpu=0, physics_gpu=0. sdk_smoke.py is an API and rendering
141
+ example. Native RTX capture can need several app updates before RGB has shape
142
+ (height, width, 4). Only input and output are host-mounted. Do not access
143
+ authentication.