Deep Spectral Methods: A Surprisingly Strong Baseline for Unsupervised Semantic Segmentation and Localization.
Luke Melas-KyriaziChristian RupprechtIro LainaAndrea VedaldiPublished in: CVPR (2022)
Keyphrases
- semantic segmentation
- spectral methods
- weakly supervised
- random walk
- manifold learning
- superpixels
- conditional random fields
- spectral clustering
- semi supervised
- unsupervised learning
- object categories
- multiscale
- natural images
- topic models
- object classes
- decision trees
- multi label
- data sets
- graphical models
- supervised learning
- information extraction
- computer vision