MedCLIPSeg: Probabilistic Vision–Language Adaptation for Data-Efficient and Generalizable Medical Image Segmentation
Overview
Medical image segmentation remains challenging due to limited annotations for training, ambiguous anatomical features, and domain shifts. While vision–language models such as CLIP offer strong cross-modal representations, their potential for dense, text-guided medical image segmentation remains underexplored. We present MedCLIPSeg, a novel… See the full description on the dataset page: https://huggingface.co/datasets/foliexxx/MedCLIPSeg.
