Unsupervised Learning of Long-Term Motion Dynamics for Videos.
Zelun LuoBoya PengDe-An HuangAlexandre AlahiLi Fei-FeiPublished in: CVPR (2017)
Keyphrases
- unsupervised learning
- long term
- periodic motion
- motion features
- space time
- video sequences
- key frames
- input video
- image sequences
- moving camera
- short term
- supervised learning
- static images
- motion capture data
- video representation
- motion cues
- dynamic scenes
- recognition of human actions
- video data
- motion analysis
- temporal domain
- motion estimation
- video frames
- camera motion
- semi supervised
- video analysis
- successive frames
- motion capture
- motion model
- motion patterns
- spatial and temporal
- motion parameters
- human motion
- video content
- dynamic textures
- video objects
- text classification
- video clips
- video stabilization
- dimensionality reduction
- traffic scenes
- optical flow
- temporal coherence
- deep architectures
- recognizing human actions
- motion segmentation
- dynamical systems
- moving objects
- training data
- object motion
- reference frame
- adjacent frames
- visual data
- action recognition
- feature points
- spatio temporal
- reinforcement learning