Text + Mask conditioned face generation on CelebA-HQ + FFHQ (test)
16.63FIDMMFace-DiT (F)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| MMFace-DiT (F)Training Objective=flow-matching2026.03 | 16.63 | 0.34 | 53 | 93.74 | 50.12 | 31.34 | 0.69 | 0.6372 | |
| MMFace-DiT (D)Training Objective=diffusion-based DDPM2026.03 | 27.95 | 0.34 | 51 | 93.95 | 49.16 | 31.69 | 0.68 | 0.6006 | |
| Unite and Conquer (UAC)2026.03 | 48.88 | 0.46 | 48 | 78.27 | 38.82 | 23.75 | 0.76 | 0.3516 | |
| Collaborative Diffusion (CD)2026.03 | 49 | 0.56 | 46 | 85.69 | 38.85 | 25.07 | 0.75 | 0.3029 | |
| ControlNet2026.03 | 49.39 | 0.57 | 41 | 82.86 | 43.95 | 25.39 | 0.75 | 0.3103 | |
| MM2Latent2026.03 | 49.78 | 0.59 | 45 | 84.57 | 38.19 | 26.78 | 0.73 | 0.3619 | |
| DDGI2026.03 | 50.88 | 0.45 | 49 | 86 | 36.02 | 24.29 | 0.76 | 0.3851 | |
| TediGAN2026.03 | 62.55 | 0.43 | 48 | 79.77 | 39.02 | 25.26 | 0.75 | 0.4061 |