POLAR: A Portrait OLAT Dataset and Generative Framework for Illumination-Aware Face Modeling

TL;DR

POLAR introduces a large-scale, physically calibrated OLAT dataset and a flow-based generative model POLARNet for realistic, controllable face relighting that preserves identity and geometry.

Contribution

The paper presents POLAR, a new extensive OLAT dataset, and POLARNet, a novel generative framework that models illumination as a continuous, physically interpretable transformation.

Findings

POLAR dataset contains over 200 subjects with 156 lighting directions.

POLARNet accurately predicts per-light responses from a single portrait.

The framework enables scalable, controllable, and realistic face relighting.

Abstract

Face relighting aims to synthesize realistic portraits under novel illumination while preserving identity and geometry. However, progress remains constrained by the limited availability of large-scale, physically consistent illumination data. To address this, we introduce POLAR, a large-scale and physically calibrated One-Light-at-a-Time (OLAT) dataset containing over 200 subjects captured under 156 lighting directions, multiple views, and diverse expressions. Building upon POLAR, we develop a flow-based generative model POLARNet that predicts per-light OLAT responses from a single portrait, capturing fine-grained and direction-aware illumination effects while preserving facial identity. Unlike diffusion or background-conditioned methods that rely on statistical or contextual cues, our formulation models illumination as a continuous, physically interpretable transformation between lighting states, enabling scalable and controllable relighting. Together, POLAR and POLARNet form a unified illumination learning framework that links real data, generative synthesis, and physically grounded relighting, establishing a self-sustaining "chicken-and-egg" cycle for scalable and reproducible portrait illumination. Our project page: https://rex0191.github.io/POLAR/.