MeSa: Masked, Geometric, and Supervised Pre-training for Monocular Depth Estimation.
Muhammad Osama KhanJunbang LiangChun-Kai WangShan YangYu LouPublished in: CoRR (2023)
Keyphrases
- depth estimation
- monocular images
- depth map
- depth cues
- stereo vision
- depth perception
- stereo matching
- depth information
- image sequences
- multi view
- scene understanding
- learning algorithm
- model based pose estimation
- super resolution
- stereo pair
- semi supervised
- real scenes
- depth estimates
- stereo images
- d scene
- dynamic scenes
- frame rate
- structure from motion
- pose estimation
- vision system
- ground truth
- high resolution
- viewpoint
- training set
- image processing
- computer vision