Xiaoyan Xing (邢晓岩)

I am a final-year ELLIS PhD candidate at the University of Amsterdam, working on computer vision and machine learning under the supervision of Prof. Dr. Theo Gevers. I am also delighted to collaborate with Prof. Anand Bhattad from JHU.
Last year, I spent a wonder"fall" time with Jon Barron's team at Google DeepMind in San Francisco, hosted by Dor Verbin.

Before my PhD, I obtained my Master's degree (MSc) from Tsinghua University and my Bachelor's degree (BEng) from Chongqing University.

I am a photography enthusiast 📸, I play the drums 🥁, and I love traveling (4 continents unlocked, 3 more to go) 🗺️.
I am open to collaborations. Feel free to drop me an email!

Email  /  LinkedIn  /  Google Scholar  /  GitHub

profile photo
News

[Aug 2025] I started my student researcher position at Google DeepMind!

[Feb 2025] LumiNet is accepted to CVPR 2025. Demo released!

[Jun 2024] I gave an invited talk at the PBDL Workshop, CVPR 2024.

[Oct 2023] I gave a contributed talk at the CV4Metaverse Workshop, ICCV 2023.

[Jul 2023] I attended ICVSS 2023 in Sicily!

[Feb 2023] I was nominated as an ELLIS PhD student!

[Sep 2022] I started my PhD adventure 🤞!

[Aug 2022] I arrived in the Netherlands!

[Jun 2022] I graduated with a Master's degree (MSc) from Tsinghua University.

[Apr 2022] I will join Delta Lab (funded by Bosch) at the University of Amsterdam as a PhD student, supervised by Prof. Dr. Theo Gevers.

[Mar 2022] One paper accepted to CVPR 2022!

[Mar 2022] One paper accepted to ICPR 2022 (early accepted)!

Research

Photography is known as the art of illumination and color—a subject in which I have a great interest. My research focuses on understanding lighting and the intrinsic properties of the physical world through images, with particular emphasis on applications such as relighting.

LumiNet: Latent Intrinsics Meets Diffusion Models for Indoor Scene Relighting
Xiaoyan Xing, Konrad Groh, Sezer Karaoglu, Theo Gevers, Anand Bhattad
CVPR 2025
arXiv / website / code / demo

Real-world indoor scene relighting without inverse rendering.

Retinex-Diffusion: On Controlling Illumination Conditions in Diffusion Models via Retinex Theory
Xiaoyan Xing, Vincent Tao Hu, Jan Hendrik Metzen, Konrad Groh, Sezer Karaoglu, Theo Gevers
Invited Talk @ PBDL Workshop, CVPR 2024
arXiv

Diffusion models "know" the illumination, and we show how to control it.

Intrinsic Appearance Decomposition Using Point Cloud Representation
Xiaoyan Xing, Konrad Groh, Sezer Karaoglu, Theo Gevers
CV4Metaverse Workshop, ICCV 2023 (Oral)
arXiv / paper / code

Point cloud-based intrinsic appearance decomposition that generalizes to unseen scenes and shapes.

Point Cloud Color Constancy
Xiaoyan Xing, Yanlin Qian, Sibo Feng, Yuhan Dong, Jiří Matas
CVPR 2022
paper / code / data

Leveraging point clouds with color and depth information to improve illumination estimation.

Dual-Illumination Weighting and Estimation
Xiaoyan Xing, Sibo Feng, Yanlin Qian, Yuhan Dong
ICPR 2022
paper

Efficient dual-illumination image synthesis for robust illumination map estimation.

Misc
tsinghua Teaching Assistant, Big Data Experiment and Application, Spring 2021
Teaching Assistant, Public Health Statistics, Spring 2020

This website is based on Jon's source code.