ZegCLIP: Towards Adapting CLIP for Zero-shot Semantic Segmentation.
Ziqin ZhouBowen ZhangYinjie LeiLingqiao LiuYifan LiuPublished in: CoRR (2022)
Keyphrases
- semantic segmentation
- object categories
- street scenes
- conditional random fields
- object detection
- superpixels
- pascal voc
- label transfer
- object classes
- scene classification
- object classification
- weakly supervised
- object recognition
- generative model
- image features
- training examples
- object class
- object segmentation
- image understanding
- image processing
- visual words
- image classification
- multi class
- image set
- image representation
- natural images
- multiscale