TagCLIP: Improving Discrimination Ability of Open-Vocabulary Semantic Segmentation.
Jingyao LiPengguang ChenShengju QianJiaya JiaPublished in: CoRR (2023)
Keyphrases
- semantic segmentation
- discrimination ability
- street scenes
- label transfer
- superpixels
- conditional random fields
- weakly supervised
- scene classification
- pascal voc
- object class
- object categories
- false positive rate
- object classes
- image understanding
- object detection
- input image
- multiscale
- computer vision
- object segmentation
- bounding box
- image set
- detection rate
- spatial information
- image representation
- multi class
- object recognition