BAPA-Net: Boundary Adaptation and Prototype Alignment for Cross-domain Semantic Segmentation.
Yahao LiuJinhong DengXinchen GaoWen LiLixin DuanPublished in: ICCV (2021)
Keyphrases
- cross domain
- semantic segmentation
- superpixels
- scene classification
- conditional random fields
- transfer learning
- weakly supervised
- knowledge transfer
- object categories
- object recognition
- text categorization
- reinforcement learning
- image classification
- data sets
- visual words
- target domain
- image understanding
- multiscale
- similarity measure
- data mining