MGMAE: Motion Guided Masking for Video Masked Autoencoding.
Bingkun HuangZhiyu ZhaoGuozhen ZhangYu QiaoLimin WangPublished in: CoRR (2023)
Keyphrases
- space time
- video sequences
- object motion
- input video
- motion features
- key frames
- spatial and temporal
- moving camera
- temporal filtering
- motion estimation
- low frame rate
- video data
- successive frames
- motion analysis
- visual data
- temporal consistency
- visual cues
- video content
- multimedia
- video frames
- shot change detection
- static images
- dynamic scenes
- motion capture data
- video objects
- temporal continuity
- human motion
- image sequences
- optical flow
- camera movement
- temporal coherence
- video representation
- dynamic textures
- motion model
- video footage
- camera motion
- video signals
- moving objects
- motion trajectories
- image frames
- stationary camera
- video streams
- temporal domain
- motion patterns
- layered representation
- action recognition
- periodic motion
- real time
- wide area motion imagery
- image motion
- video processing
- surveillance videos
- motion segmentation
- human actions
- motion field
- global motion
- video summarization
- motion capture
- video analysis
- motion detection
- motion parameters
- spatio temporal