Explore the Potential of CLIP for Training-Free Open Vocabulary Semantic Segmentation.
Tong ShaoZhuotao TianHang ZhaoJingyong SuPublished in: CoRR (2024)
Keyphrases
- semantic segmentation
- street scenes
- weakly supervised
- superpixels
- conditional random fields
- label transfer
- supervised learning
- training samples
- support vector
- hidden markov models
- object detectors
- bag of words
- pascal voc
- scene classification
- pixel level
- computer vision
- training set
- object recognition
- feature extraction