SlowFast
lastmod 2026-11-10created 2018-12-273 edits
SlowFast is a deep learning network architecture for video recognition published by Facebook AI Research. It places a Slow pathway, which captures spatial information and semantics at a low frame rate, alongside a lightweight Fast pathway, which captures fast motion at a high frame rate, and merges information between them through lateral connections. It is used as a baseline model on the Kinetics action classification and AVA action detection benchmarks, and is also released as a PyTorch codebase of the same name (PySlowFast).
See also
- Deep learning
- SlowFast
- Kinetics
- A Video Dataset of Spatio-temporally Localized Atomic Visual Actions (AVA)