Real-Time Monocular Depth Estimation Using Synthetic Data With Domain Adaptation via Image Style Transfer.
Amir Atapour AbarghoueiToby P. BreckonPublished in: CVPR (2018)
Keyphrases
- synthetic data
- depth estimation
- domain adaptation
- real time
- depth cues
- depth map
- input image
- image features
- single image
- stereo vision
- transfer learning
- scene understanding
- cross domain
- feature matching
- high resolution
- depth information
- real world
- data sets
- depth estimates
- image sequences
- image retrieval
- stereo matching
- labeled data
- image segmentation
- super resolution
- dynamic scenes
- real scenes
- similarity measure
- vision system
- prior knowledge
- lighting conditions
- optical flow
- multiple sources
- knowledge transfer
- pairwise
- object recognition
- object detection
- input data
- stereo pair
- three dimensional
- semi supervised learning
- image processing