Text-to-Image Generation on Internal Corpus SD 3.5
0.3495CLIP I-TFOCUS(Ours)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FOCUS(Ours)Fine-tuning=true, Optimal λ=true2025.10 | 0.3495 | 0.2331 | 0.5744 | 0.6383 | 22.6445 | 1.4495 | 5.9174 | |
| Divide&BindFine-tuning=true, Optimal λ=true2025.10 | 0.3486 | 0.2266 | 0.587 | 0.6358 | 22.3401 | 1.3524 | 0.8006 | |
| CONFORMFine-tuning=true, Optimal λ=true2025.10 | 0.3478 | 0.2294 | 0.5646 | 0.6393 | 22.5962 | 1.3782 | 3.4583 | |
| BaseFine-tuning=true, Optimal λ=true2025.10 | 0.3474 | 0.2309 | 0.5731 | 0.6402 | 22.694 | 1.3175 | 0 | |
| Attend&ExciteFine-tuning=true, Optimal λ=true2025.10 | 0.3469 | 0.2281 | 0.5747 | 0.6425 | 22.8429 | 1.446 | 5.7181 |