SAM-guided contrast based self-training for source-free cross-domain semantic segmentation.
Qinghua RenKe HouYongzhao ZhanChen WangPublished in: Multim. Syst. (2024)
Keyphrases
- cross domain
- semantic segmentation
- target domain
- superpixels
- scene classification
- transfer learning
- weakly supervised
- conditional random fields
- knowledge transfer
- text categorization
- object class
- object categories
- semi supervised learning
- co training
- object classes
- object detection
- unlabeled data
- image understanding
- segmentation algorithm
- e government
- semi supervised
- training set
- object recognition
- image set
- image segmentation
- feature extraction