EffiScene: Efficient Per-Pixel Rigidity Inference for Unsupervised Joint Learning of Optical Flow, Depth, Camera Pose and Motion Segmentation.
Yang JiaoGuangming ShiTrac D. TranPublished in: CoRR (2020)
Keyphrases
- motion segmentation
- optical flow
- image sequences
- moving objects
- unsupervised learning
- motion model
- motion estimation
- depth ordering
- camera pose
- flow field
- camera motion
- missing data
- model selection
- structure from motion
- motion field
- focal length
- feature tracking
- computer vision
- camera calibration
- pose estimation
- multi view
- least squares