[2111.05826] Palette: Image-to-Image Diffusion Models

This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out this https URL for an overview of the results.

This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on

related reading

• What are Diffusion Models? | Lil'Loglilianweng.github.io
• ⭐️ Diffusion Modelsandrewkchan.dev
• 1703.10593arxiv.org
• The Illustrated Stable Diffusion – Jay Alammar – Visualizing machine learning one concept at a time.jalammar.github.io
• Bare-bones Diffusion Modelsmadebyoll.in
• Yang Songyang-song.net
• [2105.05233] Diffusion Models Beat GANs on Image Synthesisarxiv.org
• [2006.11239] Denoising Diffusion Probabilistic Modelsarxiv.org
• [2112.10752] High-Resolution Image Synthesis with Latent Diffusion Modelsarxiv.org
• The Annotated Diffusion Modelhuggingface.co
• A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondencesd-complements-dino.github.io
• Unified Multimodal Models as Auto-Encodersarxiv.org