Shadow removal improves with one-step diffusion guided by reward model
ODDR: One-Step Deshadow Diffusion via Reward Guidance
Computer Vision and Pattern Recognition
Summary
Removing shadows from photos is tricky, especially when real paired examples of shadowed and shadow-free images are hard to collect. The authors introduce a method called ODDR that starts by training on fake (synthetic) shadow images to remove shadows quickly in one step. They then improve the method by teaching it with a reward model that learns to judge image quality without needing humans to label data. This makes the model work better on real photos and nearly as well as methods trained on expensive real data.
What this means in practice
- •For photo editing software developers: Add efficient shadow removal in photo editing tools that works well without large real-image datasets.
- •For autonomous vehicle perception teams: Improve object visibility by removing shadows in camera images without requiring costly annotated datasets.
Authors
Junseong Shin, Kijun Kim, Minseong Kim, Dongjin Kim, Tae Hyun Kim
Abstract
Recent advances in deep learning for shadow removal have significantly enhanced image quality and realism. However, most approaches rely on real-world paired datasets, which are costly to collect and often limited in scene diversity, leading to limited generalization. To address these limitations, we propose One-step Deshadow Diffusion via Reward guidance (ODDR), a new framework that achieves efficient and high-fidelity shadow removal without relying on real-world paired supervision. Our method begins with One-step Deshadow Diffusion (ODD), a baseline model trained on synthetic shadow data for efficient one-step shadow-free reconstruction. We further adapt ODD into ODDR using ShadowReward. In contrast to traditional, annotation-heavy approaches, ShadowReward is the first reward model for shadow removal trained entirely without human annotation. It learns to mimic human perceptual judgments by ranking synthetically generated images with controlled degradations, such as texture distortion and boundary artifacts. This reward-guided fine-tuning enables ODDR to close the synthetic-to-real domain gap. Extensive experiments show that ODD achieves strong performance without relying on real-world paired supervision, and ODDR further improves the results, narrowing the gap to fully supervised methods trained on real-world paired data while maintaining higher computational efficiency as a single-step model.