H-Net: Unsupervised Attention-based Stereo Depth Estimation Leveraging Epipolar Geometry.
Baoru HuangJian-Qing ZhengStamatia GiannarouDaniel S. ElsonPublished in: CVPR Workshops (2022)
Keyphrases
- depth estimation
- epipolar geometry
- stereo pair
- image pairs
- point correspondences
- fundamental matrix
- feature points
- depth map
- real scenes
- camera motion
- stereo camera
- geometric constraints
- disparity map
- view synthesis
- depth cues
- stereo images
- stereo vision
- stereo matching
- specular reflection
- disparity estimation
- feature matching
- semi supervised
- stereo correspondence
- image correspondences
- depth information
- super resolution
- image registration
- image matching
- d scene
- position and orientation
- closed form
- epipolar lines
- image sequences