KunhangLI commited on
Commit
b78fdff
Β·
0 Parent(s):

Triangular Resampling data release: BABEL 263D and FD2 membership

Browse files
Files changed (5) hide show
  1. .gitattributes +35 -0
  2. BABEL.zip +3 -0
  3. README.md +127 -0
  4. SHA256SUMS +2 -0
  5. manifests.zip +3 -0
.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
BABEL.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:681fdd86ebc70d24bec18b1b83e04f0070a8ae8fc3cd3a33ee7f1cd7ba2dfdcf
3
+ size 1815152669
README.md ADDED
@@ -0,0 +1,127 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: babel-amass-non-commercial
4
+ license_link: https://babel.is.tue.mpg.de/license.html
5
+ pretty_name: Triangular Resampling data release (BABEL 263D)
6
+ tags:
7
+ - motion-generation
8
+ - text-to-motion
9
+ - human-motion
10
+ - babel
11
+ - humanml3d
12
+ - diffusion
13
+ ---
14
+
15
+ # Triangular Resampling: data release
16
+
17
+ Data for **Triangular Resampling for Long-Horizon Motion Generation**
18
+ ([arXiv:2609.34697](https://arxiv.org/abs/2609.34697)).
19
+ Code: [github.com/KunhangL/TriangularResampling](https://github.com/KunhangL/TriangularResampling).
20
+
21
+ This repository holds the BABEL data and the FloodDiffusion 2 (FD2) training metadata used
22
+ in the paper. These are the exact files of the paper's runs. HumanML3D is not included; it
23
+ comes with the [FloodDiffusion release](https://huggingface.co/ShandaAI/FloodDiffusionDownloads),
24
+ whose `HumanML3D.zip` is identical to the HumanML3D copy used in the paper.
25
+
26
+ | File | Extract into | Contents | SHA-256 |
27
+ | --- | --- | --- | --- |
28
+ | `BABEL.zip` | `raw_data/` | 263D BABEL features, annotations, FD1 VAE tokens, splits | `681fdd86ebc70d24bec18b1b83e04f0070a8ae8fc3cd3a33ee7f1cd7ba2dfdcf` |
29
+ | `manifests.zip` | `raw_data/` | FD2 training membership | `71b2b95ff2c072500562aa8ebe2b42d03dbbd4102b2d07f7e9c83e566d88cc32` |
30
+
31
+ ## Download
32
+
33
+ From a clone of the code repository (checks the SHA-256 of both archives):
34
+
35
+ ```bash
36
+ python -m common.download_data # both archives into raw_data/
37
+ python -m methods.fd1.download_assets --with-dataset # FD1: deps, checkpoints, HumanML3D and BABEL
38
+ python -m methods.fd2.download_assets --with-deps --with-data # FD2: checkpoint, deps, BABEL, manifests
39
+ ```
40
+
41
+ Or directly:
42
+
43
+ ```python
44
+ import zipfile
45
+ from huggingface_hub import hf_hub_download
46
+
47
+ for name in ("BABEL.zip", "manifests.zip"):
48
+ path = hf_hub_download("KunhangLI/TriangularResampling", name)
49
+ zipfile.ZipFile(path).extractall("raw_data")
50
+ ```
51
+
52
+ ## `BABEL.zip`
53
+
54
+ ```
55
+ BABEL/
56
+ β”œβ”€β”€ new_joint_vecs/<id>.npy 5,341 sequences, float32 (T, 263), 20 fps
57
+ β”œβ”€β”€ texts/<id>.txt 4,877 annotation files
58
+ β”œβ”€β”€ TOKENS_20251030_085836_vae_wan_z4/<id>.npy 5,341 VAE token files, float32 (ceil(T/4), 4)
59
+ β”œβ”€β”€ train.txt 3,985 ids (BABEL train split)
60
+ β”œβ”€β”€ val.txt 1,356 ids (BABEL validation split)
61
+ └── test_min.txt 10 ids (validation sequences used for rendering)
62
+ ```
63
+
64
+ - **Features.** Every BABEL sequence (train and validation) in the 263-dimensional
65
+ HumanML3D motion representation at 20 fps, one file per BABEL sequence id, 25 to 5,272
66
+ frames. Sequences are complete, not cut into segments.
67
+ - **Annotations.** HumanML3D-style text files with one line per labelled BABEL segment:
68
+ `caption#word/POS ...#start_sec#end_sec`. 464 sequences have no annotation; the code
69
+ skips them.
70
+ - **Tokens.** The latents of the official FloodDiffusion VAE (`vae_1d_z4_step=300000.ckpt`),
71
+ 4 frames per token, as read by the FD1 configs. They are the tokens of the paper's FD1
72
+ runs, so FD1 can be trained without re-encoding.
73
+ - **Splits.** `train.txt` and `val.txt` follow BABEL's train / validation split. Training
74
+ uses the 3,640 annotated training sequences. The validation sequences serve as reference
75
+ motions for the paper's time-varying-text evaluation, whose frozen prompt manifest is
76
+ `common/evaluation/manifests/babel_text_streams_256.jsonl` in the code repository.
77
+
78
+ ## `manifests.zip`
79
+
80
+ ```
81
+ manifests/
82
+ └── humanml_babel.json FD2 training membership
83
+ ```
84
+
85
+ - **`humanml_babel.json`** lists every HumanML3D and BABEL sample of the FD2 run, with
86
+ paths relative to the code repository root. It has 25,966 training samples, 4,384
87
+ validation samples (the full HumanML3D test split) and 18 rendering samples. Its SHA-256,
88
+ `b27b204b48915114cd69ba016f25640e75ba809f94c496c27184cb6f63750bac`, is pinned in `methods/fd2/configs/humanml_babel_tr.yaml`. Running
89
+ `methods/fd2/prepare_data.py` from the repository root rebuilds the same file. Its
90
+ `problems` field lists two HumanML3D test features with non-finite values (`007975`,
91
+ `M007975`), which FD2 validation skips.
92
+
93
+ ## License
94
+
95
+ These files are derived from [BABEL](https://babel.is.tue.mpg.de/) (Punnakkal et al.,
96
+ CVPR 2021), which builds on [AMASS](https://amass.is.tue.mpg.de/) (Mahmood et al.,
97
+ ICCV 2019). Only extracted motion features, text annotations, model latents and id lists are
98
+ provided. No raw AMASS data (SMPL parameters or
99
+ meshes) is distributed.
100
+
101
+ The data is for **non-commercial scientific research only**. By using it you agree to the
102
+ [BABEL license](https://babel.is.tue.mpg.de/license.html) and the
103
+ [AMASS license](https://amass.is.tue.mpg.de/license.html). If you need the raw motion data or a commercial license, register
104
+ with and contact the dataset owners.
105
+
106
+ ## Citation
107
+
108
+ ```bibtex
109
+ @article{li2026triangular,
110
+ title = {Triangular Resampling for Long-Horizon Motion Generation},
111
+ author = {Li, Kunhang and Cai, Yiyi and Zhang, Xiangyue and Tu, Fangyuan and Wu, Yuhan and Wang, Zhixiang and Zhang, Kaipeng and Liu, Haiyang},
112
+ journal = {arXiv preprint arXiv:2609.34697},
113
+ year = {2026}
114
+ }
115
+ @inproceedings{punnakkal2021babel,
116
+ title = {{BABEL}: Bodies, Action and Behavior with English Labels},
117
+ author = {Punnakkal, Abhinanda R. and Chandrasekaran, Arjun and Athanasiou, Nikos and Quiros-Ramirez, Alejandra and Black, Michael J.},
118
+ booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
119
+ year = {2021}
120
+ }
121
+ @inproceedings{mahmood2019amass,
122
+ title = {{AMASS}: Archive of Motion Capture as Surface Shapes},
123
+ author = {Mahmood, Naureen and Ghorbani, Nima and Troje, Nikolaus F. and Pons-Moll, Gerard and Black, Michael J.},
124
+ booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
125
+ year = {2019}
126
+ }
127
+ ```
SHA256SUMS ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ 681fdd86ebc70d24bec18b1b83e04f0070a8ae8fc3cd3a33ee7f1cd7ba2dfdcf BABEL.zip
2
+ 71b2b95ff2c072500562aa8ebe2b42d03dbbd4102b2d07f7e9c83e566d88cc32 manifests.zip
manifests.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71b2b95ff2c072500562aa8ebe2b42d03dbbd4102b2d07f7e9c83e566d88cc32
3
+ size 8016871