Unsupervised depth prediction from monocular sequences: Improving performances through instance segmentation.
Ambroise MoreauMatei MancasThierry DutoitPublished in: CRV (2020)
Keyphrases
- relative depth
- fully unsupervised
- image segmentation
- prediction accuracy
- segmentation of natural images
- text segmentation
- image sequences
- medical images
- level set
- segmentation algorithm
- depth cues
- bayesian image segmentation
- depth ordering
- segmentation method
- unsupervised learning
- depth map
- hidden markov models
- prediction error
- sequential patterns
- image analysis
- segmentation of textured images
- pose estimation
- normalized cut
- multiscale
- edge detection
- optical flow
- independently moving objects
- pose recovery
- computer vision
- segmented images
- object segmentation
- energy function
- supervised learning