WildRelight: A Real-World Benchmark and Physics-Guided Adaptation for Single-Image Relighting

2026-05-12

AI Summary

This work addresses the limited generalization of existing single-image relighting methods in complex real-world scenes and the absence of realistic benchmarks. To this end, we introduce WildRelight, the first relighting dataset captured under diverse outdoor real-world conditions, and propose a self-supervised test-time adaptation framework that integrates physical priors with temporal consistency. Leveraging the natural temporal evolution of illumination, our method combines high dynamic range environmental lighting capture, precisely aligned time-lapse imaging, diffusion-based posterior sampling, and physically grounded lighting models to enable effective transfer from synthetic to real domains. Experiments demonstrate that our approach substantially mitigates domain shift, significantly improving both photorealism and physical plausibility of relit results on WildRelight.

Abstract

Recent single-image relighting methods, powered by advanced generative models, have achieved impressive photorealism on synthetic benchmarks. However, their effectiveness in the complex visual landscape of the real world remains largely unverified. A critical gap exists, as current datasets are typically designed for multi-view reconstruction and fail to address the unique challenges of single-image relighting. To bridge this synthetic-to-real gap, we introduce WildRelight, the first in-the-wild dataset specifically created for evaluating single-image relighting models. WildRelight features a diverse collection of high-resolution outdoor scenes, captured under strictly aligned, temporally varying natural illuminations, each paired with a high-dynamic-range environment map. Using this data, we establish a rigorous benchmark revealing that state-of-the-art models trained on synthetic data suffer from severe domain shifts. The strictly aligned temporal structure of WildRelight enables a new paradigm for domain adaptation. We demonstrate this by introducing a physics-guided inference framework that leverages the captured natural light evolution as a self-supervised constraint. By integrating Diffusion Posterior Sampling (DPS) with temporal Sampling-Aware Test-Time Adaptation (TTA), we show that the dataset allows synthetic models to align with real-world statistics on-the-fly, transforming the intractable sim-to-real challenge into a tractable self-supervised task. The dataset and code will be made publicly available to foster robust, physically-grounded relighting research.

Problem

• single-image relighting
• synthetic-to-real gap
• real-world benchmark
• domain shift
• in-the-wild dataset

Innovation

• single-image relighting
• real-world benchmark
• physics-guided adaptation
• test-time adaptation
• domain shift