Use ViTPose to do pose estimation
ViTPose paper: https://arxiv.org/abs/2204.12484 VitPose++ paper: https://arxiv.org/abs/2212.04246v3 ViTPose docs: https://huggingface.co/docs/transformers/model_doc/vitpose.
Large portions of this code just comes from ViTPose-transformers code here. This was largely to write a batch tool without gradio in the mix.
conda env create -f ./environment.yaml
conda activate pose_estimation
pip install python-magic-bin==0.4.14 # this does not work on first install. So just install on activation and it should work
python main.py -p "./tests/data/pexels-stockphotoartist-1034859.jpg"
Data is from pexels, a free-use/stock photo site.
- Add tests
- Cleanup tox