Motion Guided Token Compression for Efficient Masked Video Modeling.
Yukun FengYangming ShiFengze LiuTan YanPublished in: CoRR (2024)
Keyphrases
- space time
- video sequences
- video signals
- dynamic scenes
- spatial and temporal
- key frames
- object motion
- optical flow
- image compression
- video data
- motion capture data
- video footage
- visual cues
- compression artifacts
- temporal filtering
- temporal continuity
- long video sequences
- input video
- shot change detection
- single frame
- video objects
- surveillance videos
- visual data
- compression ratio
- motion analysis
- event detection