Motion-inductive Self-supervised Object Discovery in Videos.
Shuangrui DingWeidi XieYabo ChenRui QianXiaopeng ZhangHongkai XiongQi TianPublished in: CoRR (2022)
Keyphrases
- moving camera
- periodic motion
- motion features
- input video
- video sequences
- key frames
- image sequences
- successive frames
- camera motion
- static images
- recognition of human actions
- motion estimation
- dynamic scenes
- space time
- motion capture data
- motion analysis
- human motion
- video frames
- video stabilization
- dynamic textures
- optical flow
- motion model
- motion field
- video objects
- inductive learning
- neighboring frames
- motion trajectories
- temporal domain
- video content
- machine learning
- video analysis
- adjacent frames
- spatial and temporal
- long video sequences
- temporal segmentation
- visual motion
- motion cues
- traffic scenes
- visual data
- motion capture
- moving objects
- motion parameters
- video data
- video retrieval
- rigidly moving
- action recognition in videos
- low frame rate
- action recognition
- human activities
- human actions
- video signals
- video clips
- motion detection
- motion patterns
- reference frame
- object motion