Image Reconstruction on CelebA HQ (test)
9.54FID (Reconstruction)SODA
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| SODALatent Dim=2048, Backbone=ResNet50x22023.11 | 9.54 | — | — | — | — | — | 18.78 | 85.9 | 0.273 | — | |
| Reg-VQfeature size=16x16, codebook size=1024, resolution=256x2562023.03 | 10.09 | — | — | — | — | — | 22.05 | — | — | — | |
| Gumbel-VQfeature size=16x16, codebook size=1024, resolution=256x2562023.03 | 12.03 | — | — | — | — | — | 20.92 | — | — | — | |
| VQ-GANfeature size=16x16, codebook size=1024, resolution=256x2562023.03 | 12.71 | — | — | — | — | — | 22.44 | — | — | — | |
| TransVQCodebook Size=8912, Code Dim=642026.02 | 13.7 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCodebook Size=8912, Code Dim=642026.02 | 14.01 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCodebook Size=8912, Code Dim=642026.02 | 14.37 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCode Dimension=3, Codebook Size=89122026.02 | 16.89 | — | — | — | — | — | — | — | — | 1 | |
| TransVQCode Dimension=3, Codebook Size=89122026.02 | 17.35 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCodebook Size=1024, Code Dim=642026.02 | 17.44 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-FCCode Dimension=3, Codebook Size=89122026.02 | 17.57 | — | — | — | — | — | — | — | — | 0.34 | |
| TransVQCodebook Size=1024, Code Dim=642026.02 | 17.6 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCodebook Size=1024, Code Dim=642026.02 | 17.77 | — | — | — | — | — | — | — | — | 1 | |
| TransVQCode Dimension=3, Codebook Size=10242026.02 | 17.82 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-EMACode Dimension=3, Codebook Size=10242026.02 | 17.86 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCode Dimension=3, Codebook Size=10242026.02 | 17.86 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-EMACode Dimension=3, Codebook Size=89122026.02 | 17.93 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-FCCode Dimension=3, Codebook Size=10242026.02 | 18.48 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-EMACodebook Size=8912, Code Dim=642026.02 | 19.39 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-EMACodebook Size=1024, Code Dim=642026.02 | 20.81 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCode Dimension=3, Codebook Size=89122026.02 | 21.69 | — | — | — | — | — | — | — | — | 1 | |
| DALL-E2 (with CLIP)Latent Dim=10242023.11 | 21.91 | — | — | — | — | — | 9.34 | 31.1 | 0.484 | — | |
| TransVQCodebook Size=64, Code Dim=642026.02 | 23.24 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCode Dimension=3, Codebook Size=10242026.02 | 23.35 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCode Dimension=3, Codebook Size=642026.02 | 23.39 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCodebook Size=64, Code Dim=642026.02 | 23.39 | — | — | — | — | — | — | — | — | 1 | |
| SimVQCodebook Size=64, Code Dim=642026.02 | 23.4 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-FCCode Dimension=3, Codebook Size=642026.02 | 24.1 | — | — | — | — | — | — | — | — | 1 | |
| VQVAE2Code Dimension=3, Codebook Size=89122026.02 | 24.24 | — | — | — | — | — | — | — | — | 0.6 | |
| VQGAN-FCCodebook Size=64, Code Dim=642026.02 | 24.41 | — | — | — | — | — | — | — | — | 0.66 | |
| TransVQCode Dimension=3, Codebook Size=642026.02 | 24.42 | — | — | — | — | — | — | — | — | 1 | |
| NS-VQCode Dimension=3, Codebook Size=642026.02 | 24.55 | — | — | — | — | — | — | — | — | 1 | |
| Unconditional Diffusion2023.11 | 25.33 | — | — | — | — | — | — | — | — | — | |
| VQGAN-FCCodebook Size=1024, Code Dim=642026.02 | 25.77 | — | — | — | — | — | — | — | — | 0.03 | |
| VQVAE2Codebook Size=8912, Code Dim=642026.02 | 25.87 | — | — | — | — | — | — | — | — | 0.28 | |
| VQGAN-FCCodebook Size=8912, Code Dim=642026.02 | 26 | — | — | — | — | — | — | — | — | 0 | |
| VQ-VAEfeature size=16x16, codebook size=1024, resolution=256x2562023.03 | 28.38 | — | — | — | — | — | 23.39 | — | — | — | |
| VQVAE2Code Dimension=3, Codebook Size=10242026.02 | 31.17 | — | — | — | — | — | — | — | — | 0.4 | |
| VQVAE2Codebook Size=1024, Code Dim=642026.02 | 36.91 | — | — | — | — | — | — | — | — | 0.57 | |
| VQVAE2Code Dimension=3, Codebook Size=642026.02 | 53.25 | — | — | — | — | — | — | — | — | 1 | |
| VQGAN-EMACode Dimension=3, Codebook Size=642026.02 | 60.86 | — | — | — | — | — | — | — | — | 1 | |
| Vanilla Auto-EncoderLatent Dim=20482023.11 | 65.5 | — | — | — | — | — | 21.61 | 90.6 | 0.327 | — | |
| β-VAELatent Dim=20482023.11 | 78.65 | — | — | — | — | — | 19.94 | 88.6 | 0.314 | — | |
| B-TCVAELatent Dim=20482023.11 | 80.33 | — | — | — | — | — | 20.18 | 89.2 | 0.315 | — | |
| DIP-VAE (I)Latent Dim=20482023.11 | 81.02 | — | — | — | — | — | 19.83 | 88.4 | 0.316 | — | |
| DIP-VAE (II)Latent Dim=20482023.11 | 82.41 | — | — | — | — | — | 19.95 | 88.7 | 0.307 | — | |
| VQVAE2Codebook Size=64, Code Dim=642026.02 | 85.86 | — | — | — | — | — | — | — | — | 0.82 | |
| FactorVAELatent Dim=20482023.11 | 87.19 | — | — | — | — | — | 20.17 | 89 | 0.331 | — | |
| AnnealedVAELatent Dim=20482023.11 | 145.28 | — | — | — | — | — | 15.94 | 68.6 | 0.433 | — | |
| VQGAN-EMACodebook Size=64, Code Dim=642026.02 | 159.21 | — | — | — | — | — | — | — | — | 1 | |
| DiffAELatent Dim=5122023.11 | — | — | — | — | — | — | 15.28 | 68.1 | 0.392 | — | |
| ILO2021.06 | — | 36.15 | 40.98 | 43.46 | 47.89 | 48.21 | — | — | — | — | |
| PnPPrior Type=AWGN denoising2021.06 | — | 35.61 | 41.51 | 45.71 | 48.05 | 52.24 | — | — | — | — | |
| PnPPrior Type=artifact-removal (AR)2021.06 | — | 39.19 | 44.2 | 48.66 | 51.32 | 53.89 | — | — | — | — | |
| PULSE2021.06 | — | 27.45 | 29.98 | 33.06 | 34.25 | 34.77 | — | — | — | — | |
| REDPrior Type=AWGN denoising2021.06 | — | 35.46 | 41.59 | 45.65 | 48.13 | 52.17 | — | — | — | — | |
| StyleGAN2, WLatent Dim=5122023.11 | — | — | — | — | — | — | 16.76 | 66.2 | 0.394 | — | |
| StyleGAN2, W+Latent Dim=512x142023.11 | — | — | — | — | — | — | 21.42 | 81.3 | 0.345 | — | |
| TV2021.06 | — | 32.13 | 35.24 | 37.41 | 39.35 | 41.29 | — | — | — | — | |
| VQGANLatent Dim=256x16x162023.11 | — | — | — | — | — | — | 23.28 | 77.3 | 0.311 | — |