DEVIAS: Learning Disentangled Video Representations of Action and Scene for Holistic Video Understanding.
Kyungho BaeGeo AhnYoungrae KimJinwoo ChoiPublished in: CoRR (2023)
Keyphrases
- video sequences
- video streams
- learning algorithm
- video frames
- interactive video
- video content
- multimedia
- video data
- video clips
- external representations
- video database
- dynamic scenes
- space time
- human actions
- learning process
- spatial and temporal
- video footage
- real time
- visual data
- input video
- online learning
- action selection
- scene change detection
- live video
- action models
- motion features
- video images
- moving camera
- video retrieval
- key frames
- video analysis
- cognitive processing
- action recognition
- object detection and tracking
- reinforcement learning