ComfyUI Dwpose TensorRT

This project provides a Tensorrt implementation of Dwpose for ultra fast pose estimation inside ComfyUI

This project is licensed under CC BY-NC-SA, everyone is FREE to access, use, modify and redistribute with the same license.

For commercial purposes, please contact me directly at yuvraj108c@gmail.com

If you like the project, please give me a star! ⭐

⏱️ Performance

Note: The following results were benchmarked on FP16 engines inside ComfyUI, using 1000 similar frames

| Device | FPS | | :----: | :-: | | L40s | 20 |

🚀 Installation

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

🛠️ Building Tensorrt Engine

1. Download the following onnx models:

   • dw-ll_ucoco_384.onnx
   • yolox_l.onnx

2. Build tensorrt engines for both of these models by running:

   • python export_trt.py

3. Place the exported engines inside ComfyUI /models/tensorrt/dwpose directory

☀️ Usage

• Insert node by Right Click -> tensorrt -> Dwpose Tensorrt

🤖 Environment tested

• Ubuntu 22.04 LTS, Cuda 12.4, Tensorrt 10.2.0.post1, Python 3.10, L40s GPU
• Windows (Not tested, but should work)

👏 Credits

• https://github.com/IDEA-Research/DWPose
• https://github.com/legraphista/dwpose-video

License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)