SlowFast Networks for Video Recognition

TL;DR

SlowFast networks utilize dual pathways at different frame rates to effectively capture spatial semantics and motion details, achieving state-of-the-art results in video recognition tasks.

Contribution

Introduction of the SlowFast architecture, combining a low frame rate pathway with a high frame rate pathway for improved video understanding.

Findings

Achieved state-of-the-art accuracy on Kinetics, Charades, and AVA datasets.

Fast pathway can be lightweight yet effective for motion capture.

Significant performance improvements over previous models.

Abstract

We present SlowFast networks for video recognition. Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution. The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition. Our models achieve strong performance for both action classification and detection in video, and large improvements are pin-pointed as contributions by our SlowFast concept. We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA. Code has been made available at: https://github.com/facebookresearch/SlowFast