Neural Marionette: Unsupervised Learning of Motion Skeleton and Latent Dynamics from Volumetric Video.
Jinseok BaeHojun JangCheol-Hui MinHyungun ChoiYoung Min KimPublished in: AAAI (2022)
Keyphrases
- unsupervised learning
- space time
- articulated human motion
- temporal filtering
- video sequences
- object motion
- key frames
- moving camera
- motion features
- spatial and temporal
- successive frames
- input video
- video content
- motion capture
- image sequences
- motion model
- motion estimation
- visual data
- static images
- temporal consistency
- video frames
- video objects
- motion trajectories
- low frame rate
- motion analysis
- supervised learning
- camera motion
- human motion
- dynamic scenes
- optical flow
- video data
- visual cues
- dynamic textures
- video streams
- neural network
- multimedia
- semi supervised
- motion capture data
- network architecture
- motion patterns
- temporal continuity
- dynamical systems
- moving objects
- machine learning
- layered representation
- video shots
- motion field
- latent variables
- spatio temporal
- pairwise
- video analysis
- principal component analysis
- dimensionality reduction
- object recognition
- feature extraction
- neural learning
- three dimensional
- dynamic model
- human actions