RTMDet: Optimized for Qualcomm Devices
RTMDet is a highly efficient model for real-time object detection,capable of predicting both the bounding boxes and classes of objects within an image.It is highly optimized for real-time applications, making it reliable for industrial and commercial use
This is based on the implementation of RTMDet found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.
Getting Started
Due to licensing restrictions, we cannot distribute pre-exported model assets for this model. Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations
See our repository for RTMDet on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.object_detection
Model Stats:
- Input resolution: 640x640
- Model checkpoint: RTMDet Medium
- Model size (float): 105 MB
- Number of parameters: 27.5M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| RTMDet | ONNX | float | Snapdragon® X2 Elite | 13.261 ms | 5 - 5 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® X Elite | 16.53 ms | 51 - 51 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 12.417 ms | 1 - 239 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 39.587 ms | 3 - 300 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 26.845 ms | 5 - 13 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 16.277 ms | 5 - 7 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® QCS8450 | 39.587 ms | 3 - 300 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 24.25 ms | 5 - 12 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 16.53 ms | 51 - 51 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 9.496 ms | 3 - 193 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Elite Mobile | 9.496 ms | 3 - 193 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 7.196 ms | 2 - 197 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® X2 Elite | 6.688 ms | 2 - 2 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® X Elite | 15.701 ms | 27 - 27 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 9.947 ms | 3 - 357 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 20.843 ms | 3 - 359 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 65.71 ms | 25 - 28 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 14.004 ms | 2 - 6 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.237 ms | 2 - 5 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® QCS8450 | 20.843 ms | 3 - 359 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 15.715 ms | 2 - 5 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 15.701 ms | 27 - 27 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 161.586 ms | 28 - 347 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 22.915 ms | 25 - 353 MB | NPU |
| RTMDet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 7.251 ms | 0 - 300 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 7.251 ms | 0 - 300 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 5.913 ms | 1 - 328 MB | NPU |
| RTMDet | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 22.915 ms | 25 - 353 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 12.126 ms | 9 - 299 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 46.761 ms | 9 - 359 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 27.508 ms | 9 - 81 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 85.514 ms | 10 - 221 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.991 ms | 9 - 12 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8775P | 24.256 ms | 9 - 223 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8650P | 24.256 ms | 9 - 223 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8255P | 24.256 ms | 9 - 223 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® QCS8450 | 46.761 ms | 9 - 359 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 24.514 ms | 8 - 79 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 9.126 ms | 9 - 226 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA7255P | 85.514 ms | 10 - 221 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8295P | 33.048 ms | 9 - 282 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Elite Mobile | 9.126 ms | 9 - 226 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.781 ms | 8 - 223 MB | NPU |
License
- The license for the original implementation of RTMDet can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.
