Random epipolar constraint loss functions for supervised optical flow estimation.
Zhengyuan FanZemin CaiPublished in: Pattern Recognit. (2024)
Keyphrases
- loss function
- optical flow estimation
- multi frame
- optical flow
- flow field
- pairwise
- camera motion
- coarse to fine
- motion field
- motion model
- epipolar geometry
- support vector
- stereo matching
- semi supervised
- motion estimation
- multiscale
- point correspondences
- image pairs
- optic flow
- rigid motion
- image sequences
- computationally efficient
- super resolution
- least squares
- machine learning
- motion analysis
- image matching
- feature matching
- d scene
- structure from motion
- feature detection
- image analysis
- velocity field
- feature selection
- smoothness constraint