Scene Graph Detection on VG 1.0 (test)
34.5R@50VS3
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VS3Backbone=Swin-L2024.04 | 34.5 | — | — | — | — | 39.2 | |
| SVRPBackbone=ResNet-502024.04 | 31.8 | — | — | 10.5 | 12.8 | 35.8 | |
| ISGBackbone=ResNet-1012024.04 | 29.7 | 2.7 | 3.8 | 8 | 8.8 | 32.1 | |
| SGTRBackbone=ResNet-1012024.04 | 24.6 | 2.5 | 5.8 | 12 | 15.2 | 28.4 | |
| SSRCNNBackbone=X101-FPN2024.04 | 23.7 | 3.1 | 4.5 | 18.6 | 22.5 | 27.3 | |
| RelTRBackbone=ResNet-1012024.04 | 21.2 | 2.6 | 3.4 | 6.8 | 10.8 | 27.5 | |
| PGSGBackbone=ViT-B (BLIP), Classifier=close-set2024.04 | 20.3 | 4.8 | 7.6 | 10.5 | 12.7 | 23.6 | |
| SGTRBackbone=ViT-B (BLIP)2024.04 | 18.1 | 2.1 | 4.9 | 8.7 | 10.1 | 20.4 | |
| PGSGBackbone=ViT-B (BLIP)2024.04 | 16.7 | 6.2 | 8.5 | 8.9 | 11.5 | 21.2 |