- This research is conducted at VISHC
pip install dtw
pip install opencv-python
conda install --file requirements.txt
python train.py
- If you want to use different parts of the loss or change training settings, simply modify the
config.py
.
Here we provide a sample evaluation procedure
python evaluate_KT.py --logdir /path/to/trained/model/
- Note that a useful function for framewise feature extraction is provided under
utils.py
The data processing, loading and training setup code was modified from this very useful repo: https://github.com/google-research/google-research/tree/master/tcc