Metric: FID (lower is better)
| # | Model↕ | FID▲ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Palatte (20-30% free form) | 11.7 | No | Palette: Image-to-Image Diffusion Models | 2021-11-10 | Code |
| 2 | Palette (128×128 center mask) | 11.9 | No | Palette: Image-to-Image Diffusion Models | 2021-11-10 | Code |
| 3 | DeepFillv2 (20-30% free form) | 13.5 | No | Free-Form Image Inpainting with Gated Convolution | 2018-06-10 | Code |
| 4 | TFill (20-30% free form) | 15.2 | No | Bridging Global Context Interactions for High-Fi... | 2021-04-02 | Code |
| 5 | DeepFillv2 (128×128 center mask) | 15.3 | No | Free-Form Image Inpainting with Gated Convolution | 2018-06-10 | Code |
| 6 | HiFill (20-30% free form) | 15.7 | No | Contextual Residual Aggregation for Ultra High-R... | 2020-05-19 | Code |
| 7 | HiFill (128×128 center mask) | 16.9 | No | Contextual Residual Aggregation for Ultra High-R... | 2020-05-19 | Code |