Unsupervised Semantic Segmentation by Distilling Feature Correspondences.
Mark HamiltonZhoutong ZhangBharath HariharanNoah SnavelyWilliam T. FreemanPublished in: CoRR (2022)
Keyphrases
- semantic segmentation
- feature correspondences
- weakly supervised
- multiple views
- structure from motion
- coarse to fine
- superpixels
- bundle adjustment
- camera calibration
- conditional random fields
- semi supervised
- scene classification
- object class
- motion parameters
- multiple images
- geometric constraints
- object categories
- object classes
- multi view
- multiresolution
- viewpoint
- object recognition
- training data
- computer vision
- least squares