POLAR：面向光照感知人脸建模的肖像 OLAT 数据集与生成框架
POLAR: A Portrait OLAT Dataset and Generative Framework for Illumination-Aware Face Modeling
用 flow-based 生成模型从单张肖像预测逐光源 OLAT 响应，把光照建模为光照状态间的连续可解释变换，为可控重打光提供了不依赖扩散先验、物理可解释的替代范式，其 OLAT 数据与生成框架的闭环思路也可迁移到其他逆渲染任务。
原文依据 we develop a flow-based generative model POLARNet that predicts per-light OLAT responses from a single portrait
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论人脸重打光旨在在新光照条件下合成逼真的肖像，同时保持身份与几何结构。然而，大规模、物理一致的光照数据的稀缺限制了该方向的进展。为此，我们提出 POLAR，一个大规模且经过物理标定的逐光源（One-Light-at-a-Time, OLAT）数据集，包含 200 余名受试者，在 156 个光照方向、多视角和多种表情下采集。基于 POLAR，我们构建了基于流的生成模型 POLARNet，它能够从单张肖像预测逐光源的 OLAT 响应，在保持人脸身份的同时捕捉细粒度、方向感知的光照效果。与依赖统计或上下文线索的扩散模型或背景条件方法不同，我们的形式将光照建模为光照状态之间连续且物理可解释的变换，从而实现可扩展且可控的重打光。POLAR 与 POLARNet 共同构成一个统一的光照学习框架，将真实数据、生成式合成与物理可解释的重打光联系起来，为可扩展且可复现的肖像光照建立了一个自维持的“先有鸡还是先有蛋”循环。
英文原文摘要 1221 字符
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.