TAFormer: A Unified Target-Aware Transformer for Video and Motion Joint Prediction in Aerial Scenes.
Liangyu XuWanxuan LuHongfeng YuYongqiang MaoHanbo BiChenglong LiuXian SunKun FuPublished in: IEEE Trans. Geosci. Remote. Sens. (2024)
Keyphrases
- dynamic scenes
- video footage
- space time
- video scene
- dynamic textures
- video sequences
- video data
- moving camera
- object motion
- input video
- key frames
- temporal filtering
- moving target
- spatial and temporal
- video frames
- successive frames
- static images
- surveillance videos
- temporal continuity
- visual data
- human motion
- visual cues
- motion trajectories
- independently moving objects
- motion segmentation
- motion model
- target object
- prediction error
- moving objects
- low frame rate
- video objects
- motion features
- motion capture data
- dynamic background
- image sequences
- video analysis
- symbolic descriptions
- video clips
- interesting events
- motion estimation
- live video
- layered representation
- man made
- image motion
- motion analysis
- motion patterns
- multimedia
- multi view
- motion vectors
- scalable video coding
- video summarization
- motion field
- video surveillance
- motion parameters
- video streams
- fault diagnosis
- background subtraction
- fuzzy logic
- robust tracking
- optical flow
- spatio temporal
- motion detection