RePaint: Inpainting using Denoising Diffusion Probabilistic Models

TLDR

Free‑form inpainting adds content to arbitrary masked regions, but existing methods trained on specific mask distributions and pixel‑wise losses often produce only textural extensions rather than semantically meaningful results. This paper introduces RePaint, a diffusion‑based inpainting method that works even with extreme masks. RePaint leverages a pretrained unconditional DDPM, altering only the reverse diffusion steps by sampling unmasked pixels from the input image, without modifying the network, and is evaluated on faces and general images with standard and extreme masks. RePaint generates high‑quality, diverse outputs and outperforms state‑of‑the‑art autoregressive and GAN approaches on at least five of six mask distributions. Code is available at git.io/RePaint.

Abstract

Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint