Image Colorization on ImageNet (val)
14.68FIDGround truth
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Ground truth2021.02 | 14.68 | — | — | — | — | — | — | — | |
| ColTran2021.02 | 19.37 | — | — | — | — | — | — | — | |
| ColTran-BArchitecture variant=Axial Transformer with conditioning via addition2021.02 | 19.98 | — | — | — | — | — | — | — | |
| ColTran-STraining Resolution=28x282021.02 | 22.06 | — | — | — | — | — | — | — | |
| PixColor2021.02 | 24.32 | — | — | — | — | — | — | — | |
| cGAN2021.02 | 24.41 | — | — | — | — | — | — | — | |
| cINN2021.02 | 25.13 | — | — | — | — | — | — | — | |
| VAE-MDN2021.02 | 25.98 | — | — | — | — | — | — | — | |
| Grayscale2021.02 | 30.19 | — | — | — | — | — | — | — | |
| BEITPre-training=IN-21k2022.09 | — | — | — | — | — | — | 1.25 | 0.73 | |
| BEITPre-training=Figures dataset2022.09 | — | — | — | — | — | — | 0.6 | 0.7 | |
| Coltran2021.11 | — | 19.37 | — | — | — | 36.55 | — | — | |
| COLTRAN2022.05 | — | 19.37 | — | — | — | — | — | — | |
| Copybaseline=true2022.09 | — | — | — | — | — | — | 2.63 | 0.75 | |
| MAEPre-training=IN-1k2022.09 | — | — | — | — | — | — | 1.13 | 0.87 | |
| MAEPre-training=Figures dataset2022.09 | — | — | — | — | — | — | 0.43 | 0.55 | |
| MAE-VQGANPre-training=IN-1k2022.09 | — | — | — | — | — | — | 3.31 | 0.75 | |
| MAE-VQGANPre-training=Figures dataset2022.09 | — | — | — | — | — | — | 0.67 | 0.4 | |
| Original images2021.11 | — | 14.68 | 229.6 | 75.6 | 0 | — | — | — | |
| Palette2021.11 | — | 15.78 | 200.8 | 72.5 | 46.2 | 47.8 | — | — | |
| Palette2022.05 | — | 15.78 | — | — | — | — | — | — | |
| pix2pix2021.11 | — | 24.41 | — | — | — | 29.9 | — | — | |
| PixColor2021.11 | — | 24.32 | — | — | — | — | — | — | |
| Regression2021.11 | — | 17.89 | 169.8 | 68.2 | 60 | 39.45 | — | — | |
| UViMBackbone=ViT-L/16 (Encoder), ViT-B (Decoder), Pre-training=ImageNet-21k, Input Resolution=512x5122022.05 | — | 16.99 | — | — | — | — | — | — | |
| VQGANPre-training=IN-1k2022.09 | — | — | — | — | — | — | 2.44 | 0.66 | |
| VQGANPre-training=Figures dataset2022.09 | — | — | — | — | — | — | 1.5 | 0.56 |