Motion-Augmented Self-Training for Video Recognition at Smaller Scale.
Kirill GavrilyukMihir JainIlia KarmanovCees G. M. SnoekPublished in: CoRR (2021)
Keyphrases
- space time
- static images
- video sequences
- low frame rate
- motion history images
- motion features
- spatial and temporal
- input video
- key frames
- recognition rate
- gait recognition
- temporal filtering
- video data
- object motion
- motion estimation
- human motion
- moving camera
- human activities
- activity recognition
- visual perception
- object recognition
- successive frames
- dynamic scenes
- motion recognition
- image sequences
- shot change detection
- motion capture data
- visual data
- motion analysis
- human gait
- temporal segmentation
- video frames
- hand gestures
- visual cues
- video streams
- recognition algorithm
- action recognition
- temporal consistency
- temporal coherence
- motion model
- surveillance videos
- video content
- dynamic textures
- motion trajectories
- moving objects
- multimedia
- training set
- optical flow
- semi supervised learning
- motion field
- camera motion
- motion parameters
- human actions
- co training
- temporal continuity
- feature extraction
- layered representation
- video scene
- video analysis
- video signals
- scale space
- video clips
- video objects
- motion capture