Depth Anywhere: Enhancing 360 Monocular Depth Estimation via Perspective Distillation and Unlabeled Data Augmentation.
Ning-Hsu WangYu-Lun LiuPublished in: CoRR (2024)
Keyphrases
- depth estimation
- unlabeled data
- depth cues
- labeled data
- semi supervised learning
- depth map
- stereo vision
- semi supervised
- depth information
- depth estimates
- active learning
- stereo matching
- training data
- text classification
- dynamic scenes
- multi view
- scene understanding
- image sequences
- supervised learning
- real scenes
- learning algorithm
- training set
- data points
- stereo pair
- unsupervised learning
- d scene
- viewpoint
- super resolution
- prior knowledge
- training examples
- feature matching
- three dimensional
- machine learning
- high resolution
- object recognition
- depth from defocus
- dynamic programming
- multi class
- pairwise
- disparity map
- low resolution
- input image
- specular reflection
- decision trees