MGMAE: Motion Guided Masking for Video Masked Autoencoding.
Bingkun HuangZhiyu ZhaoGuozhen ZhangYu QiaoLimin WangPublished in: ICCV (2023)
Keyphrases
- space time
- spatial and temporal
- video sequences
- object motion
- key frames
- temporal filtering
- successive frames
- shot change detection
- video data
- input video
- motion estimation
- motion features
- visual data
- low frame rate
- image sequences
- motion analysis
- temporal domain
- temporal consistency
- video representation
- dynamic scenes
- video footage
- temporal continuity
- optical flow
- image frames
- moving camera
- video frames
- camera motion
- static images
- video analysis
- visual cues
- video content
- video objects
- video streams
- motion patterns
- motion detection
- temporal coherence
- dynamic textures
- video scene
- real time
- motion trajectories
- single frame
- video processing
- surveillance videos
- video database
- motion capture data
- reference frame
- video retrieval
- motion field
- spatio temporal
- computational complexity
- multiview video
- multimedia
- action recognition
- human motion
- video signals
- visual information
- human activities
- wide area motion imagery
- moving objects
- multi view
- periodic motion
- perceptual distortion
- motion vectors
- video surveillance
- motion parameters
- humanoid robot
- video shots
- temporal structure
- image motion