All Tools
S
OtherFreeOpen Source
SLOWFAST
FAIR video backbones for classification and detection
Apache-2.0
ABOUT
Video research needs a shared PyTorch stack for SlowFast, X3D, and MViT instead of one-off training scripts. PySlowFast ships those backbones with Kinetics and AVA recipes so labs can train, fine-tune, and detect actions from a single FAIR codebase.
INTEGRATION GUIDE
1. Train SlowFast or X3D classifiers on Kinetics with the official configs
2. Run AVA action detection from a PySlowFast checkpoint
3. Pretrain video transformers with MaskFeat, MAE, or MViTv2 projects
4. Benchmark a new video backbone against the PySlowFast model zoo
TAGS
pythonvideopytorchclassificationdetectionfairtransformers