This repository hosts the prepared dataset package used by Concept-to-Pixel (C2P), a prompt-free universal medical image segmentation framework associated with the ECCV 2026 paper:
Concept-to-Pixel: Prompt-Free Universal Medical Image Segmentation
The repository is intended as a ready-to-use training package for the official C2P codebase rather than a single cleaned benchmark export. In addition to images and masks, it… See the full description on the dataset page: https://huggingface.co/datasets/yunyundi/c2p-dataset.