S4T: Source-free domain adaptation for semantic segmentation via self-supervised selective self-training.
Viraj PrabhuShivam KhareDeeksha KartikJudy HoffmanPublished in: CoRR (2021)
Keyphrases
- domain adaptation
- semantic segmentation
- multiple sources
- co training
- target domain
- semi supervised learning
- semi supervised
- labeled data
- superpixels
- conditional random fields
- transfer learning
- cross domain
- unlabeled data
- scene classification
- training data
- object categories
- active learning
- test data
- sentiment classification
- domain specific
- object classes
- supervised learning
- training set
- training samples
- text classification
- high dimensional
- data sets
- multi view
- multiscale
- feature selection
- computer vision
- machine learning