Grounded Image Generation on LVIS (val)
25.2APUpper-bound
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Upper-boundPre-training data=–, Training data=–2023.01 | 25.2 | — | 19 | 22.2 | 31.2 | |
| GLIGEN-LDMPre-training data=GoldG,O365,SBU,CC3M, Training data=LVIS2023.01 | 14.9 | 6.25 | 10.1 | 12.8 | 19.3 | |
| GLIGEN-LDMPre-training data=GoldG,O365,SBU,CC3M, Training data=–2023.01 | 11.1 | 10.28 | 9 | 9.8 | 13.4 | |
| GLIGEN-LDMPre-training data=GoldG,O365, Training data=–2023.01 | 10.6 | 8.45 | 5.8 | 9.6 | 13.8 | |
| GLIGEN-LDMPre-training data=COCO2014CD, Training data=–2023.01 | 6.4 | 22.17 | 5.8 | 5.8 | 7.4 | |
| GLIGEN-LDMPre-training data=COCO2014G, Training data=–2023.01 | 6 | 13.48 | 4.4 | 6.1 | 6.6 | |
| GLIGEN-LDMPre-training data=COCO2014D, Training data=–2023.01 | 4.4 | 31.31 | 2.3 | 3.3 | 6.5 | |
| LAMAPre-training data=–, Training data=LVIS2023.01 | 2 | 151.96 | 0.9 | 1.3 | 3.2 |