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:
- Model checkpoint: RTMDet Medium
- Input resolution: 640x640
- Number of parameters: 27.5M
- Model size (float): 105 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| RTMDet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.96 ms | 5 - 189 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® X2 Elite | 8.177 ms | 53 - 53 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® X Elite | 14.238 ms | 51 - 51 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 10.625 ms | 2 - 232 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 13.63 ms | 0 - 54 MB | NPU |
| RTMDet | ONNX | float | Qualcomm® QCS9075 | 23.805 ms | 5 - 12 MB | NPU |
| RTMDet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 8.295 ms | 3 - 184 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 10.284 ms | 3 - 330 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Snapdragon® X2 Elite | 11.075 ms | 32 - 32 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Snapdragon® X Elite | 29.491 ms | 29 - 29 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 19.725 ms | 3 - 383 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Qualcomm® QCS8550 (Proxy) | 27.975 ms | 0 - 35 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Qualcomm® QCS9075 | 33.566 ms | 2 - 5 MB | NPU |
| RTMDet | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Elite For Galaxy Mobile | 14.299 ms | 1 - 313 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.812 ms | 0 - 210 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 11.719 ms | 0 - 285 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 84.066 ms | 0 - 206 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 15.822 ms | 0 - 3 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8775P | 22.999 ms | 0 - 207 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® QCS9075 | 24.351 ms | 0 - 62 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 37.541 ms | 0 - 347 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA7255P | 84.066 ms | 0 - 206 MB | NPU |
| RTMDet | TFLITE | float | Qualcomm® SA8295P | 29.934 ms | 0 - 267 MB | NPU |
| RTMDet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 9.133 ms | 0 - 213 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.