Spatiotemporal Initialization for 3D CNNs with Generated Motion Patterns.
Hirokatsu KataokaKensho HaraRyusuke HayashiEisuke YamagataNakamasa InouePublished in: WACV (2022)
Keyphrases
- motion patterns
- space time
- spatio temporal
- moving objects
- spatial and temporal
- image sequences
- motion analysis
- motion model
- dynamic textures
- video sequences
- activity analysis
- cellular neural networks
- dynamic scenes
- appearance model
- visual motion
- motion trajectories
- bayesian framework
- hand gestures
- reinforcement learning
- similar patterns
- bayesian networks
- viewpoint
- human actions
- three dimensional
- multiple views
- feature space
- multi view
- low cost
- dynamic programming
- optical flow