MotionZero: Exploiting Motion Priors for Zero-shot Text-to-Video Generation.
Sitong SuLitao GuoLianli GaoHengtao ShenJingkuan SongPublished in: CoRR (2023)
Keyphrases
- space time
- temporal filtering
- spatial and temporal
- video sequences
- key frames
- object motion
- moving camera
- motion features
- text generation
- motion estimation
- image sequences
- motion analysis
- input video
- video scene
- dynamic scenes
- successive frames
- shot change detection
- low frame rate
- video data
- visual cues
- visual data
- video segments
- optical flow
- natural language descriptions
- temporal continuity
- video objects
- news video
- surveillance videos
- video footage
- multimedia
- video content
- temporal coherence
- motion model
- video frames
- camera motion
- static images
- video search
- motion trajectories
- video analysis
- video streams
- text detection
- temporal consistency
- moving objects
- human motion
- layered representation
- text mining
- motion segmentation
- motion capture
- single frame
- video database
- reference frame
- video summarization
- information retrieval
- image frames
- dynamic textures
- camera movement
- motion patterns
- wide area motion imagery
- motion capture data
- multimedia documents
- video shots
- video clips
- video retrieval
- video surveillance
- event detection
- temporal information