Saved in:
Bibliographic Details
Main Authors: Huang, Qian, Jaiswal, Mayoore Selvarasa, Zhong, Zhen, Pereira, Rochelle, Min, Jianyuan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.23094
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908993183547392
author Huang, Qian
Jaiswal, Mayoore Selvarasa
Zhong, Zhen
Pereira, Rochelle
Min, Jianyuan
author_facet Huang, Qian
Jaiswal, Mayoore Selvarasa
Zhong, Zhen
Pereira, Rochelle
Min, Jianyuan
contents The real-world adoption of portrait relighting is hindered by dataset domain gaps, camera sensitivity, and computational costs. We address these challenges with Hybrid Domain Knowledge Fusion, a paradigm that fuses the specialized strengths of synthetic, One-Light-at-A-Time (OLAT), and real-world datasets into a compact model. Our approach features specialized prior models hardened by domain-aware adaptation, followed by augmented knowledge distillation into a lightweight student model with multi-domain expertise. Our method demonstrates a 6x to 240x inference speedup while maintaining state-of-the-art (SOTA) visual quality in the experiments. Additionally, we construct a massive, high-fidelity synthetic dataset with diverse ground-truth intrinsics to support our training pipeline.
format Preprint
id arxiv_https___arxiv_org_abs_2604_23094
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Toward Real-World Adoption of Portrait Relighting via Hybrid Domain Knowledge Fusion
Huang, Qian
Jaiswal, Mayoore Selvarasa
Zhong, Zhen
Pereira, Rochelle
Min, Jianyuan
Computer Vision and Pattern Recognition
Graphics
Machine Learning
The real-world adoption of portrait relighting is hindered by dataset domain gaps, camera sensitivity, and computational costs. We address these challenges with Hybrid Domain Knowledge Fusion, a paradigm that fuses the specialized strengths of synthetic, One-Light-at-A-Time (OLAT), and real-world datasets into a compact model. Our approach features specialized prior models hardened by domain-aware adaptation, followed by augmented knowledge distillation into a lightweight student model with multi-domain expertise. Our method demonstrates a 6x to 240x inference speedup while maintaining state-of-the-art (SOTA) visual quality in the experiments. Additionally, we construct a massive, high-fidelity synthetic dataset with diverse ground-truth intrinsics to support our training pipeline.
title Toward Real-World Adoption of Portrait Relighting via Hybrid Domain Knowledge Fusion
topic Computer Vision and Pattern Recognition
Graphics
Machine Learning
url https://arxiv.org/abs/2604.23094