This repository currently implemented the CPM model using TensorFlow for hand pose estimation. The original source is here

I did small changes in the network to train hand pose estimation model.

Demo

1. Install python packages (requirements.txt)

pip install -r requirements.txt

2. Compile MobileNet-Yolo. This is for hand detection. I've trained hand detection caffe model.. Please reference the MobileNet-Yolo README file to compile and run the framework. Another option is use your own hand detection method.
3. Run the demo

python demo.py

Example of result shown in result.mp4

Training

You should prepare your dataset, and change the path in the hand.cfg file.

cd training
python3 src/train.py experiments/hand.cfg

Project is updating............, If you have idea to improve the result please open an issue and share your thoughts