Box-to-image generation on COCO (val)
50.2AP@50 (Box)Upper bound (real images)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Upper bound (real images)Location format input=Boxes2024.02 | 50.2 | 66.7 | 61 | — | |
| InstanceDiffusion (hybrid)Location format input=Boxes, Conditioning type=hybrid2024.02 | 44.6 | 59.6 | 58.8 | 25.5 | |
| InstanceDiffusionLocation format input=Boxes2024.02 | 38.8 | 55.4 | 52.9 | 23.9 | |
| GLIGENLocation format input=Boxes2024.02 | 19.6 | 35 | 30.7 | 27 | |
| GLIGENLocation format input=Boxes, Recognition model=YOLOv82024.02 | 19.3 | 34.6 | 31.1 | — |