MeSa: Masked, Geometric, and Supervised Pre-training for Monocular Depth Estimation.
Muhammad Osama KhanJunbang LiangChun-Kai WangShan YangYu LouPublished in: UniReps (2023)
Keyphrases
- depth estimation
- monocular images
- depth cues
- depth map
- stereo vision
- depth perception
- stereo matching
- image sequences
- real scenes
- feature matching
- depth estimates
- depth information
- scene understanding
- stereo pair
- semi supervised
- model based pose estimation
- learning algorithm
- ground truth
- image processing
- dynamic scenes
- geometric constraints
- stereo images
- d scene
- pose estimation
- super resolution
- specular reflection
- multi view
- training set
- high quality
- real time