This project provides a Tensorrt implementation of Dwpose for ultra fast pose estimation inside ComfyUI
If you like my projects and wish to see updates and new features, please consider supporting me. It helps a lot!
Note: The following results were benchmarked inside ComfyUI, using 1000 similar frames
| Device | FPS |
|---|---|
| L40s | 20 |
| RTX4090 | 18.7 |
Navigate to the ComfyUI /custom_nodes directory
git clone https://github.com/yuvraj108c/ComfyUI-Dwpose-Tensorrt
cd ./ComfyUI-Dwpose-Tensorrt
pip install -r requirements.txt- Insert node by
Right Click -> tensorrt -> Dwpose Tensorrt - Tensorrt engines will be built automatically
- Set DWPOSE precision to FP32 for better accuracy
- Automatic tensorrt engines are built from the workflow itself, to simplify the process for non-technical people
- Ubuntu 22.04 LTS, Cuda 12.4, Tensorrt 10.2.0.post1, Python 3.10, L40s GPU
- Windows (Not tested, but should work)
- https://github.com/IDEA-Research/DWPose
- https://github.com/legraphista/dwpose-video
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)