Unsupervised depth estimation from monocular videos with hybrid geometric-refined loss and contextual attention.
Mingliang ZhangXinchen YeXin FanWei ZhongPublished in: Neurocomputing (2020)
Keyphrases
- depth estimation
- monocular images
- dynamic scenes
- depth cues
- stereo vision
- depth map
- stereo matching
- depth information
- video sequences
- depth estimates
- scene understanding
- multi view
- image sequences
- real scenes
- super resolution
- stereo pair
- moving camera
- geometric constraints
- video surveillance
- semi supervised
- disparity map
- video frames
- feature matching
- vision system
- motion estimation
- space time
- pose estimation
- lighting conditions