Parsing All Adverse Scenes: Severity-Aware Semantic Segmentation with Mask-Enhanced Cross-Domain Consistency.
Fuhao LiZiyang GongYupeng DengXianzheng MaRenrui ZhangZhenming JiXiangwei ZhuHong ZhangPublished in: AAAI (2024)
Keyphrases
- cross domain
- semantic segmentation
- superpixels
- conditional random fields
- scene classification
- knowledge transfer
- transfer learning
- weakly supervised
- natural scenes
- object categories
- natural language processing
- object classes
- computer vision
- natural language
- multiscale
- bounding box
- image understanding
- text categorization
- e government
- object recognition
- object detection
- multiple objects
- training examples
- d objects
- support vector
- natural images