ResearchBenchmarksSemantic Segmentation on VisASemantic Segmentation on VisAFollow36.23mIoUUniDG-RFT24.540427.575230.6133.6448Apr 10, 2026Evaluation resultsMethodMethodLinksmIoUmIoU (Background)mIoU (Foreground)UniDG-RFTEvaluation protocol=Tr...Evaluation protocol=Training-Free Few-shot Image Insertion2026.0436.2399.5523.47UniDG-SFTEvaluation protocol=Tr...Evaluation protocol=Training-Free Few-shot Image Insertion2026.0431.7999.5418.89SeaSEvaluation protocol=Fe...Evaluation protocol=Few-shot Anomaly Generation2026.0424.9999.5811.83