Multi-body SE(3) Equivariance for Unsupervised Rigid Segmentation and Motion Estimation.
Jia-Xing ZhongTa Ying ChengYuhang HeKai LuKaichen ZhouAndrew MarkhamNiki TrigoniPublished in: CoRR (2023)
Keyphrases
- multibody
- motion segmentation
- motion estimation
- motion model
- image sequences
- dynamic scenes
- optical flow
- motion tracking
- geometric properties
- motion field
- motion vectors
- flow field
- video coding
- motion compensation
- video sequences
- unsupervised learning
- super resolution
- semi supervised
- moving objects
- point correspondences
- structure from motion
- missing data
- motion parameters
- multi frame
- feature tracking
- model selection
- articulated objects
- computational complexity
- iterative algorithms