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