CLIP-DINOiser: Enhancing CLIP for Open-Vocabulary Semantic Segmentation

Enhancing CLIP features for open-vocabulary semantic segmentation without annotations.