TAP-Vid: A Benchmark for Tracking Any Point in a Video.
Carl DoerschAnkush GuptaLarisa MarkeevaAdrià RecasensLucas SmairaYusuf AytarJoão CarreiraAndrew ZissermanYi YangPublished in: NeurIPS (2022)
Keyphrases
- face detection and tracking
- object segmentation and tracking
- real time
- video images
- object detection and tracking
- person detection
- video surveillance
- low frame rate
- temporal continuity
- video data
- video sequences
- video analysis
- object tracking
- video dataset
- video content
- surveillance videos
- multimedia
- crowded scenes
- object motion
- image frames
- video database
- stationary camera
- video tracking
- appearance model
- moving camera
- video streams
- input video
- articulated human motion
- space time
- motion model
- kalman filter
- successive frames
- video scene
- text detection
- dynamic background
- live video
- video frames
- motion segmentation
- spatial and temporal resolution
- video clips
- point tracking
- particle filter
- image sequences
- background subtraction
- visual tracking
- video shots
- wide area motion imagery