SF-Net: Single-Frame Supervision for Temporal Action Localization.
Fan MaLinchao ZhuYi YangShengxin ZhaGourab KunduMatt FeiszliZheng ShouPublished in: CoRR (2020)
Keyphrases
- single frame
- temporal redundancy
- multi frame
- multiple frames
- video sequences
- super resolution
- image sequences
- optical flow
- spatio temporal
- low resolution
- high resolution video
- video frames
- temporal coherence
- temporal information
- surveillance videos
- human actions
- motion detection
- space time
- motion estimation
- motion field
- computer vision
- object recognition
- three dimensional