To setup your environment, make sure that you have a NVIDIA GPU with support for CUDA 12.1+.
conda env create -f conda_env.yamlconda activate torch2.2
Then, read the documentation on the datasets to download, prepare and generate annotation files to be able to train the models.
See python experiment.py --help for help on how to run the experiments. For example configuration, see the scripts/slurm/example.slurm.
See python extract_keyframes.py --help for help on how to extract keyframes.