Unsupervised Video Action Clustering via Motion-Scene Interaction Constraint.
Bo PengJianjun LeiHuazhu FuChangqing ZhangTat-Seng ChuaXuelong LiPublished in: IEEE Trans. Circuits Syst. Video Technol. (2020)
Keyphrases
- video sequences
- input video
- motion features
- object motion
- dynamic scenes
- space time
- moving camera
- video scene
- human actions
- recognizing human actions
- unsupervised learning
- unsupervised manner
- image sequences
- single frame
- multiple frames
- moving objects
- visual data
- video footage
- video data
- camera motion
- static images
- image motion
- video frames
- action recognition
- layered representation
- dynamic textures
- surveillance videos
- stationary camera
- clustering algorithm
- human motion
- motion estimation
- consecutive frames
- object detection and tracking
- action classification
- spatial and temporal
- k means
- image frames
- motion analysis
- camera movement
- d scene
- motion segmentation
- video images
- video content
- optical flow
- motion model
- key frames
- motion patterns
- body motions
- motion trajectories
- smoothness constraint
- crowded scenes
- background subtraction
- motion history images
- video surveillance
- high frame rate
- activity recognition
- neighboring frames
- spatio temporal
- video objects
- visual features
- single image
- dynamic background
- motion parameters
- video shots
- traffic scenes
- video streams
- ground plane
- atomic actions
- ego motion
- computer vision
- human activities
- affine camera
- spectral clustering
- multi view
- point features
- visual input