Improved Noise and Attack Robustness for Semantic Segmentation by Using Multi-Task Training with Self-Supervised Depth Estimation.
Marvin KlingnerAndreas BärTim FingscheidtPublished in: CVPR Workshops (2020)
Keyphrases
- semantic segmentation
- depth estimation
- stereo vision
- depth map
- superpixels
- stereo matching
- conditional random fields
- scene understanding
- depth information
- dynamic scenes
- training set
- real scenes
- object categories
- feature matching
- feature selection
- vision system
- supervised learning
- multi class
- disparity map
- object recognition