Scene-Graph-Based Image Retrieval on Flickr30k (test)
60.01NDCG@1PRISm
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PRISmModality=Graph2025.12 | 60.01 | 62.84 | 64.08 | 61.25 | 67.25 | 70.15 | 74.35 | |
| Hi-SIGIRModality=Graph2025.12 | 56.49 | 57.36 | 59.27 | 50.36 | 52.59 | 47.52 | 56.15 | |
| CLIPModality=Vision-Language2025.12 | 52.47 | 52.59 | 52.4 | 60.59 | 68.6 | 67.79 | 72.59 | |
| BLIP-2Modality=Vision-Language2025.12 | 52.34 | 52.17 | 51.87 | 60.23 | 68.92 | 67.57 | 72.57 | |
| DINOv3Modality=Visual2025.12 | 51.05 | 51.89 | 51.64 | 58.16 | 67.9 | 67.36 | 70.08 | |
| MambaVisionModality=Visual2025.12 | 50.32 | 50.8 | 50.78 | 56.2 | 64.2 | 60.24 | 68.76 | |
| IRSGSModality=Graph2025.12 | 30.35 | 31.43 | 31.73 | 28.75 | 37.42 | 38.67 | 43.77 | |
| SCENIRModality=Graph2025.12 | 28.41 | 28.84 | 29.14 | 25.52 | 18.55 | 15.28 | 40.42 |