Scene Graph Detection on Visual Genome (VG) (test)
18.6mR@50Structured Sparse R-CNN
Evaluation results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Structured Sparse R-CNNdebiasing=LA, queries=8002021.06 | 18.6 | 18.2 | 23.7 | 27.3 | 2 | 3.1 | 4.5 | 13.7 | 22.5 | 0.32 | — | — | — | — | — | — | |
| Structured Sparse R-CNNdebiasing=LA, queries=standard2021.06 | 17.9 | 18.4 | 23.3 | 26.5 | 1.9 | 2.9 | 4 | 13.5 | 21.4 | 0.19 | — | — | — | — | — | — | |
| GPSnet + ADTransBackbone=ResNet-502023.07 | 15.8 | — | — | — | — | — | — | 12.3 | 19.2 | — | 15.1 | 18.6 | 21.9 | — | — | — | |
| MOTIFS + ADTransBackbone=ResNet-502023.07 | 15.5 | — | — | — | — | — | — | 10.6 | 18.1 | — | 13.4 | 18.9 | 22 | — | — | — | |
| Structured Sparse R-CNNdebiasing=TDE, queries=standard2021.06 | 15 | 14.5 | 18.3 | 21 | 1.8 | 2.7 | 3.6 | 10.8 | 18.5 | 0.29 | — | — | — | — | — | — | |
| Structured Sparse R-CNNdebiasing=TDE, queries=8002021.06 | 14.6 | 15 | 19.7 | 22.9 | 1.6 | 2.7 | 3.8 | 9.8 | 18 | 0.54 | — | — | — | — | — | — | |
| VCTree + ADTransBackbone=ResNet-502023.07 | 12.5 | — | — | — | — | — | — | 9.7 | 16.9 | — | 12.2 | 16.3 | 20.3 | — | — | — | |
| BGNNdebiasing=none2021.06 | 10.7 | 31 | 35.8 | — | — | — | — | — | 12.6 | — | — | — | — | — | — | — | |
| VCTree(EBM)debiasing=TDE2021.06 | 9.7 | 14.7 | 20.6 | 24.7 | 1.6 | 2.7 | — | 7.1 | 11.6 | — | — | — | — | — | — | — | |
| VCTreedebiasing=TDE2021.06 | 9.3 | 14 | 19.4 | 23.2 | — | 2.6 | 3.2 | 6.9 | 11.1 | — | — | — | — | — | — | — | |
| Transformerdebiasing=TDE2021.06 | 9.2 | 11.2 | 15.6 | 19 | 1.4 | 2 | 2.5 | 6.9 | 10.9 | — | — | — | — | — | — | — | |
| VTransEdebiasing=TDE2021.06 | 8.6 | 13.5 | 18.7 | 22.6 | 2 | 2.7 | — | 6.3 | 10.5 | — | — | — | — | — | — | — | |
| Structured Sparse R-CNNdebiasing=none, queries=8002021.06 | 8.6 | 26.1 | 33.5 | 38.4 | 1.5 | 2.7 | 4 | 6.2 | 10.3 | 0.32 | — | — | — | — | — | — | |
| Structured Sparse R-CNNdebiasing=none, queries=standard2021.06 | 8.4 | 25.8 | 32.7 | 36.9 | 1.5 | 2.7 | 3.7 | 6.1 | 10 | 0.19 | — | — | — | — | — | — | |
| MOTIFSdebiasing=TDE2021.06 | 8.2 | 12.4 | 16.9 | 20.3 | — | 2.3 | 2.9 | 5.8 | 9.8 | — | — | — | — | — | — | — | |
| Transformerdebiasing=none2021.06 | 8.1 | 25.6 | 33 | 37.4 | 0 | 0.1 | 0.3 | 6 | 9.6 | 0.38 | — | — | — | — | — | — | |
| MOTIFS(EBM)debiasing=none2021.06 | 7.7 | 24.3 | 31.7 | 36.3 | — | 0.1 | 0.2 | 5.7 | 9.3 | — | — | — | — | — | — | — | |
| VCTree(EBM)debiasing=none2021.06 | 7.7 | 24.2 | 31.4 | 35.9 | — | 0.2 | 0.4 | 5.7 | 9.1 | — | — | — | — | — | — | — | |
| VCTreedebiasing=none2021.06 | 7.4 | 24.5 | 31.9 | 36.2 | 0.1 | 0.3 | 0.7 | 5.4 | 8.7 | 0.67 | — | — | — | — | — | — | |
| GPS-Netdebiasing=none2021.06 | 7 | — | 31.1 | 35.9 | — | — | — | — | 8.6 | — | — | — | — | — | — | — | |
| VTransEdebiasing=none2021.06 | 6.8 | 24.5 | 31.3 | 35.5 | — | 1.9 | 2.6 | 5.1 | 8 | 0.4 | — | — | — | — | — | — | |
| VCTreeBackbone=ResNet-502023.07 | 6.7 | — | — | — | — | — | — | 5.2 | 7.9 | — | 8.7 | 11 | 13 | — | — | — | |
| MOTIFSBackbone=ResNet-502023.07 | 6.2 | — | — | — | — | — | — | 4.8 | 7.1 | — | 8 | 10.3 | 11.8 | — | — | — | |
| RelDNdebiasing=none2021.06 | 6 | — | 31.4 | 35.9 | — | — | — | — | 7.3 | — | — | — | — | — | — | — | |
| GPSnetBackbone=ResNet-502023.07 | 5.9 | — | — | — | — | — | — | 5.2 | 7.1 | — | 8.6 | 9.9 | 11.8 | — | — | — | |
| G-RCNNdebiasing=none2021.06 | 5.8 | — | 29.7 | 32.8 | — | — | — | — | 6.6 | — | — | — | — | — | — | — | |
| MOTIFSdebiasing=none2021.06 | 5.5 | 25.1 | 32.1 | 36.9 | — | 0.1 | 0.2 | 4.1 | 6.8 | 0.45 | — | — | — | — | — | — | |
| MOTIFS(Focal)debiasing=none2021.06 | 5.3 | 24.7 | 31.7 | 36.7 | — | 0.1 | 0.3 | 3.9 | 6.6 | — | — | — | — | — | — | — | |
| IMPdebiasing=none2021.06 | 4.2 | 18.1 | 25.9 | 31.2 | 0.2 | 0.4 | 0.8 | 2.8 | 5.3 | 0.43 | — | — | — | — | — | — | |
| DPLBackbone=Motifs2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 10.1 | 15.3 | 20 | |
| DPLBackbone=VCTree2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 10.6 | 16.3 | 21.1 | |
| MotifsBackbone=Motifs2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 7.5 | 12.5 | 16.9 | |
| PCPLBackbone=Motifs2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 9.8 | 14.8 | 19.6 | |
| PCPLBackbone=VCTree2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 9.9 | 15.1 | 19.9 | |
| RwtBackbone=Motifs2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 8 | 12.9 | 16.8 | |
| RwtBackbone=VCTree2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 8 | 12.1 | 15.9 | |
| TDEBackbone=Motifs2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 7.4 | 11.2 | 14.9 | |
| TDEBackbone=VCTree2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 8 | 12.9 | 16.8 | |
| VCTreeBackbone=VCTree2024.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | 7.5 | 12.5 | 16.7 |