Semi-supervised semantic segmentation needs strong, varied perturbations.
Geoffrey FrenchSamuli LaineTimo AilaMichal MackiewiczGraham D. FinlaysonPublished in: BMVC (2020)
Keyphrases
- semantic segmentation
- semi supervised
- weakly supervised
- supervised methods
- street scenes
- conditional random fields
- superpixels
- label transfer
- semi supervised learning
- scene classification
- pascal voc
- pairwise
- object categories
- active learning
- object class
- object classes
- multi view
- object detection
- image understanding
- unlabeled data
- labeled data
- object recognition
- supervised learning
- segmentation algorithm
- multiscale
- bayesian networks
- relation extraction
- image set
- action recognition
- generative model