Train One, Generalize to All: Generalizable Semantic Segmentation from Single-Scene to All Adverse Scenes.
Ziyang GongFuhao LiYupeng DengWenjun ShenXianzheng MaZhenming JiNan XiaPublished in: ACM Multimedia (2023)
Keyphrases
- semantic segmentation
- scene classification
- natural scenes
- superpixels
- label transfer
- real scenes
- dynamic scenes
- video sequences
- multiple objects
- multiple images
- weakly supervised
- image set
- single image
- natural images
- three dimensional
- conditional random fields
- keypoints
- image understanding
- segmentation algorithm
- input image