Grounded Text-to-Image Generation on LVIS (val)
25.2APUpper-bound
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Upper-bound2023.01 | 25.2 | 19 | 22.2 | 31.2 | |
| GLIGEN-LDMTraining data=GoldG,O365,SBU,CC3M2023.01 | 11.1 | 9 | 9.8 | 13.4 | |
| GLIGEN-StableTraining data=GoldG,O365,SBU,CC3M2023.01 | 10.8 | 8.8 | 9.9 | 12.6 | |
| GLIGEN-LDMTraining data=GoldG,O3652023.01 | 10.6 | 5.8 | 9.6 | 13.8 | |
| GLIGEN-LDMTraining data=COCO2014CD2023.01 | 6.4 | 5.8 | 5.8 | 7.4 | |
| GLIGEN-LDMTraining data=COCO2014G2023.01 | 6 | 4.4 | 6.1 | 6.6 | |
| GLIGEN-LDMTraining data=COCO2014D2023.01 | 4.4 | 2.3 | 3.3 | 6.5 | |
| LAMATraining data=LVIS2023.01 | 2 | 0.9 | 1.3 | 3.2 |