| --- |
| license: apache-2.0 |
| tags: |
| - diffusion |
| - pixel-diffusion |
| - text-to-image |
| - image-generation |
| - imagenet |
| pipeline_tag: unconditional-image-generation |
| --- |
| |
| # DiP: Taming Diffusion Models in Pixel Space |
|
|
| <div style="text-align: center;"> |
| <a href="https://arxiv.org/abs/2511.18822"><img src="https://img.shields.io/badge/arXiv-2511.18822-b31b1b.svg" alt="arXiv"></a> |
| </div> |
|
|
| ## Introduction |
| Diffusion models face a fundamental trade-off between generation quality and computational efficiency. Latent Diffusion Models (LDMs) offer an efficient solution but suffer from potential information loss and non-end-to-end training. In contrast, existing pixel space models bypass VAEs but are computationally prohibitive for high-resolution synthesis. To resolve this dilemma, we propose DiP, an efficient pixel space diffusion framework. DiP decouples generation into a global and a local stage: a Diffusion Transformer (DiT) backbone operates on large patches for efficient global structure construction, while a co-trained lightweight Patch Detailer Head leverages contextual features to restore fine-grained local details. This synergistic design achieves computational efficiency comparable to LDMs without relying on a VAE. DiP is accomplished with up to 10x faster inference speeds than previous method while increasing the total number of parameters by only 0.3%, and achieves an 1.79 FID score on ImageNet 256x256. |
|
|
