AlignZeg: Mitigating Objective Misalignment for Zero-shot Semantic Segmentation.
Jiannan GeLingxi XieHongtao XiePandeng LiXiaopeng ZhangYongdong ZhangQi TianPublished in: CoRR (2024)
Keyphrases
- semantic segmentation
- object categories
- street scenes
- superpixels
- conditional random fields
- label transfer
- weakly supervised
- pascal voc
- scene classification
- object classes
- object class
- object detection
- object recognition
- object classification
- visual words
- training examples
- object segmentation
- image classification
- pixel level
- feature extraction