SEGIC: Unleashing the Emergent Correspondence for In-Context Segmentation

We introduce SEGIC, an end-to-end segment-in-context framework built upon a single frozen vision foundation model.

Model framework