← back to paper
arxiv: 2605.05077 · 2 revisions
FlowDIS: Language-Guided Dichotomous Image Segmentation with Flow Matching