New Approach Reframes Cross-Domain Few-Shot Segmentation via Reference Inversion
A new arXiv paper proposes reference inversion for cross-domain few-shot segmentation, shifting from visual matching to textual conditioning to segment objects in unseen domains with few annotated references.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-04 05:00 · arXiv cs.CV
From Visual Matching to Textual Conditioning: Cross-Domain Few-Shot Segmentation via Reference Inversion