Geometry-Aware Symmetric Domain Adaptation for Monocular Depth Estimation.
Shanshan ZhaoHuan FuMingming GongDacheng TaoPublished in: CVPR (2019)
Keyphrases
- domain adaptation
- depth estimation
- stereo vision
- depth map
- stereo matching
- depth information
- multiple sources
- image sequences
- three dimensional
- labeled data
- cross domain
- stereo pair
- dynamic scenes
- scene understanding
- semi supervised
- super resolution
- real scenes
- semi supervised learning
- disparity map
- feature matching
- multi view
- test data
- optical flow
- pose estimation
- unlabeled data
- sentiment classification
- transfer learning
- target domain
- specular reflection
- active learning
- geometric constraints
- stereo images
- training data
- image pairs
- data sets
- high quality
- unsupervised learning
- dynamic programming