Scene Graph Detection on Panoptic Scene Graph (PSG) (test)
40.6R@20HiLo
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HiLoBackbone=Swin-L2023.03 | 40.6 | 29.7 | 48.7 | 37.6 | 51.4 | 40.9 | |
| VLPromptModel Input=Vision + Language2023.11 | 39.4 | 34.7 | 47.6 | 45.1 | 52.4 | 53.7 | |
| HiLoBackbone=Swin-B2023.03 | 38.5 | 28.3 | 46.2 | 35.3 | 49.6 | 39.1 | |
| HiLoBackbone=R502023.03 | 34.1 | 23.7 | 40.7 | 30.3 | 43 | 33.1 | |
| HiLoModel Input=Vision2023.11 | 34.1 | 23.7 | 40.7 | 30.3 | 43 | 33.1 | |
| PairNetModel Input=Vision2023.11 | 29.6 | 24.7 | 35.6 | 28.5 | 39.6 | 30.6 | |
| PSGTRBackbone=R502023.03 | 28.4 | 16.6 | 34.4 | 20.8 | 36.3 | 22.1 | |
| PSGTRModel Input=Vision2023.11 | 28.4 | 16.6 | 34.4 | 20.8 | 36.3 | 22.1 | |
| VCTreeBackbone=R502023.03 | 20.6 | 9.7 | 22.1 | 10.2 | 22.5 | 10.2 | |
| VCTreeModel Input=Vision2023.11 | 20.6 | 9.7 | 22.1 | 10.2 | 22.5 | 10.2 | |
| MOTIFBackbone=R502023.03 | 20 | 9.1 | 21.7 | 9.6 | 22 | 9.7 | |
| MOTIFModel Input=Vision2023.11 | 20 | 9.1 | 21.7 | 9.6 | 22 | 9.7 | |
| PSGFormerBackbone=R502023.03 | 18 | 14.8 | 19.6 | 17 | 20.1 | 17.6 | |
| PSGFormerModel Input=Vision2023.11 | 18 | 14.8 | 19.6 | 17 | 20.1 | 17.6 | |
| GPSNetBackbone=R502023.03 | 17.8 | 7 | 19.6 | 7.5 | 20.1 | 7.7 | |
| GPSNetModel Input=Vision2023.11 | 17.8 | 7 | 19.6 | 7.5 | 20.1 | 7.7 | |
| IMPBackbone=R502023.03 | 16.5 | 6.5 | 18.2 | 7.1 | 18.6 | 7.2 | |
| IMPModel Input=Vision2023.11 | 16.5 | 6.5 | 18.2 | 7.1 | 18.6 | 7.2 |