MOSO: Decomposing MOtion, Scene and Object for Video Prediction.
Mingzhen SunWeining WangXinxin ZhuJing LiuPublished in: CoRR (2023)
Keyphrases
- object motion
- video scene
- moving objects
- video sequences
- object detection and tracking
- dynamic scenes
- camera movement
- image sequences
- input video
- fixed camera
- video data
- space time
- moving camera
- video objects
- independently moving objects
- stationary camera
- camera motion
- rigid body motion
- motion features
- motion trajectories
- image frames
- surveillance videos
- consecutive frames
- multiple objects
- target object
- layered representation
- global motion
- visual input
- camera images
- rigid objects
- successive frames
- visual data
- object trajectories
- video surveillance
- video frames
- video analysis
- object tracking
- ground plane
- video footage
- optical flow
- multiple cameras
- motion parameters
- combining information from multiple
- video shots
- image motion
- temporal continuity
- background subtraction
- human motion
- object appearance
- visual scene
- dynamic textures
- single frame
- static images
- complex scenes
- object detection
- motion segmentation
- key frames
- relative position
- foreground objects
- motion patterns
- d scene
- three dimensional
- motion model
- relative depth
- video streams
- object segmentation
- foreground regions