This repository serves as tools for using the ap10k model and others from the ViTPose
Here, all you will need is the weights and configs downloaded from ViTPose
Make sure to do the following, as the packages are required to run the keypoints script:
cd mmcv
pip install -e .
cd ..
cd ViTPose
pip install -e .
Certain models are grouped with several models being contained in one .pth file, ('+' in the .pth file) you will need to split the model in its parts. To split the model, you will need to run the following command:
python ViTPose/tools/split_model.py --source /path/to/model.pth
To get and inference keypoints from your model, simply update the pose_config and pose_model in the keypoints.py file.
python keypoints.py path/to/img