Training command is as follows:

python main.py --modality [MODALITY] --data_dir [DATA_DIR] --fusion_mode [FUSION_MODE]

DATA_DIR is the directory where you put the RGB and Optical Flow frames. MODALITY is one of RGB, Flow, Joint to train models for different modalities. FUSION_MODE is only used for Joint modality, it is one of cat, avg, mlp that corresponds to fusion by feautre concatenation, averaging or multi-layer perception learning. The model weights would be saved under logs/[MODALITY]_[NUM_SEGMENTS]_[FUSION_MODE]

Testing commad is as follows:

python test.py --modality [MODALITY] --data_dir [DATA_DIR] --fusion_mode [FUSION_MODE]

where it will output the mean and std for the classficiation accuracies of 600 randomly selected tasks.