MaXTron: Mask Transformer with Trajectory Attention for Video Panoptic Segmentation.
Ju HeQihang YuInkyu ShinXueqing DengXiaohui ShenAlan L. YuilleLiang-Chieh ChenPublished in: CoRR (2023)
Keyphrases
- video segmentation
- video sequences
- object segmentation and tracking
- video scene
- point trajectories
- segmentation algorithm
- feature trajectories
- motion segmentation
- level set
- video data
- segmentation method
- video streams
- region growing
- foreground background segmentation
- objects in video sequences
- video objects
- image segmentation
- medical images
- story segmentation
- video frames
- energy function
- key frames
- fully automatic
- video clips
- object segmentation
- single frame
- segmentation accuracy
- fault diagnosis
- video database
- bounding box
- temporal segmentation
- spatial and temporal
- optical flow
- video surveillance
- power transformers
- high voltage
- shape prior
- motion trajectories
- video shots
- video analysis
- real time
- visual attention
- event detection
- space time
- fuzzy logic
- multiscale
- image sequences
- image processing