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

TL;DR

WildRelight introduces a real-world dataset and a physics-guided adaptation method for single-image relighting, addressing the domain gap between synthetic training data and real outdoor scenes.

Contribution

The paper presents WildRelight, the first in-the-wild dataset for single-image relighting, and a physics-guided inference framework that enables models to adapt to real-world lighting conditions.

Findings

State-of-the-art models trained on synthetic data perform poorly on real scenes.

The proposed physics-guided adaptation improves relighting quality on WildRelight.

The dataset and method facilitate self-supervised domain adaptation for relighting.

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.