Inconsistency and intrinsic difficulty of different diffusion-based generative AI models. Classification error rate and obtained with Dual Vision Transformer —DaViT— model.
| Kandinsky 2.2 | JPEG Compression 90 | JPEG Compression 80 | JPEG Compression 70 | JPEG Compression 60 | Upsampling 10% | Downsampling 10% | Denoising | Sharpening | Dernoising then sharpening | |
|---|---|---|---|---|---|---|---|---|---|---|
| Holystic | 0.94% | 2.44% | 2.93% | 3.01% | 3.59% | 1.43% | 1.60% | 1.86% | 1.44% | 2.72% |
| Atomistic | 0.94% | 2.44% | 3.46% | 4.07% | 4.32% | 1.54% | 1.84% | 1.50% | 1.61% | 1.80% |
| Weighted | 0.94% | 1.42% | 2.33% | 2.79% | 3.32% | 1.12% | 1.31% | 1.68% | 1.07% | 1.68% |
| Kandinsky 2.2 | JPEG Compression 90 | JPEG Compression 80 | JPEG Compression 70 | JPEG Compression 60 | Upsampling 10% | Downsampling 10% | Denoising | Sharpening | Dernoising then sharpening | |
|---|---|---|---|---|---|---|---|---|---|---|
| Holystic | 0.94% | 2.44% | 2.93% | 3.01% | 3.59% | 1.43% | 1.60% | 1.86% | 1.44% | 2.72% |
| Atomistic | 0.94% | 2.44% | 3.46% | 4.07% | 4.32% | 1.54% | 1.84% | 1.50% | 1.61% | 1.80% |
| Weighted | 0.94% | 1.42% | 2.33% | 2.79% | 3.32% | 1.12% | 1.31% | 1.68% | 1.07% | 1.68% |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.