Text-to-Image Generation on MS-COCO 2014 (test)
0.2603CLIP ScoreKandinsky-2.2
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Kandinsky-2.2alpha=0.1, image size=512 x 512, inference steps=50, guidance scale=7.02026.05 | 0.2603 | 0.0062 | -0.2541 | 0.4407 | -0.1404 | -0.5811 | |
| SD3.5-Lalpha=0.1, image size=512 x 512, inference steps=50, guidance scale=7.02026.05 | 0.2596 | 0.0116 | -0.248 | 0.4531 | -0.1365 | -0.5896 | |
| SD3-Malpha=0.1, image size=512 x 512, inference steps=50, guidance scale=7.02026.05 | 0.259 | 0.0134 | -0.2456 | 0.4615 | -0.148 | -0.6095 | |
| FLUX.1-devalpha=0.1, image size=512 x 512, inference steps=50, guidance scale=7.02026.05 | 0.2509 | 0.0056 | -0.2453 | 0.4395 | -0.1411 | -0.5806 |