Human keypoint-grounded image generation on COCO 2017 (val)
62.4APUpper-bound
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Upper-boundSource=real images, Input resolution=256x2562023.01 | 62.4 | — | 75 | 65.9 | |
| GLIGENCaption usage=without caption, Input resolution=256x2562023.01 | 31.8 | 31.02 | 53.5 | 31 | |
| GLIGENCaption usage=with caption, Input resolution=256x2562023.01 | 31.5 | 27.34 | 52.9 | 31 | |
| pix2pixHDInput resolution=256x2562023.01 | 15.8 | 142.4 | 33.7 | 13 |