A Semi-Supervised Approach to Monocular Depth Estimation, Depth Refinement, and Semantic Segmentation of Driving Scenes using a Siamese Triple Decoder Architecture.
John Paul Tan YusiongProspero C. NavalPublished in: Informatica (Slovenia) (2020)
Keyphrases
- depth estimation
- street scenes
- dynamic scenes
- semi supervised
- real scenes
- depth map
- semantic segmentation
- multi view
- stereo vision
- depth information
- stereo matching
- scene understanding
- image sequences
- weakly supervised
- stereo pair
- super resolution
- moving objects
- space time
- motion segmentation
- d scene
- vision system
- conditional random fields
- stereo images
- pose estimation
- video sequences
- feature matching
- viewpoint
- active learning
- object detection
- three dimensional
- multiscale
- scene classification
- specular reflection
- input image
- machine learning
- superpixels
- disparity map
- matching algorithm