Multi-concept grounding on Multi-Concept Confusion Dataset
68.85mIoU (fg)AnchorDiff
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| AnchorDiffBackbone=SD3.52026.05 | 68.85 | 89.38 | 12.21 | 89.97 | |
| AnchorDiffBackbone=SD32026.05 | 64.13 | 87.27 | 16.86 | 86.24 | |
| Seg4DiffBackbone=SD3.52026.05 | 46.57 | 84.22 | 19.08 | 69.19 | |
| Seg4DiffBackbone=SD32026.05 | 44.56 | 82.06 | 22.47 | 65.88 | |
| ConceptAttentionBackbone=SD3.52026.05 | 39.97 | 88.73 | 26.36 | 59.28 | |
| ConceptAttentionBackbone=SD32026.05 | 38.24 | 80.63 | 32.95 | 55.26 | |
| Cross AttentionBackbone=SD32026.05 | 33.93 | 83.7 | 32.49 | 54.62 | |
| Cross AttentionBackbone=SD3.52026.05 | 28.37 | 88.06 | 28.45 | 65.97 | |
| DAAMBackbone=SD22026.05 | 27.1 | 71.03 | 44.08 | 47.02 | |
| OVAM2026.05 | 26.28 | 61.76 | 49.58 | 45.76 | |
| CLIPasRNN2026.05 | 25.82 | 74.89 | 47.2 | 46.81 | |
| TextSpan2026.05 | 25.47 | 59.53 | 45.99 | 52.38 | |
| DAAMBackbone=SDXL2026.05 | 24.46 | 64.12 | 50.66 | 45.31 |