WarpI2I: Image Warping for Image-to-Image Translation
For practitioners of image-to-image translation using latent diffusion models, this method offers a simple, model-agnostic way to improve fine-grained detail preservation without architectural changes.
WarpI2I proposes a saliency-guided warp-unwarp framework for image-to-image translation that reallocates latent space representation to salient regions, improving structural preservation without increasing latent resolution. Experiments on human relighting, driving scene translation, and video show improved structural preservation and lighting faithfulness with negligible overhead.
Image-to-image (I2I) translation has achieved strong results in tasks like human relighting and driving scene translation using latent diffusion models (LDMs). However, compact LDMs often struggle to preserve fine-grained structures because the encoder compresses high-resolution inputs into a spatially downsampled latent space. To address this issue, we propose a simple saliency-guided warp-unwarp framework that reallocates spatial representation toward salient regions before encoding, enabling better preservation of structural details without increasing latent resolution. The warped image is processed by the original diffusion model and then mapped back via an inverse warp. In addition, we propose a simple and efficient outpainting-based synthetic data generation pipeline to produce high-quality paired data for image relighting. Our method is model-agnostic, requires no architectural modification, and introduces negligible computational overhead. Experiments on human relighting, driving scene relighting, and translation demonstrate improved structural preservation, lighting faithfulness, and image quality, with our framework extending naturally to video via frame-by-frame application with good temporal stability. Project Webpage: https://shenzheng2000.github.io/WarpI2I.github.io