Implementation of "X3D: Expanding Architectures for Efficient Video Recognition" [ArXiv] based on the original repository.
This model has no dependencies beyond the standard dependencies found in ~/setup.py.
Four pretrained models were supplied for X3D, denotes by their sizes (XS, S, M L). These can by downloaded by executing
cd weights
./download_weights.sh