Zitian Zhang, 张梓添

I am a PhD candidate in computer science at Université Laval. I feel very fortunate to have the opportunity to work with Prof. Jean-François Lalonde and this amazing lab! My research focuses on understanding, representing, and manipulating lighting with generative models.

I was an intern research scientist at Adobe Research London in the summer of 2025, working on lighting representations and image relighting with Valentin Deschaintre, Iliyan Georgiev, Michael Fischer, and Yannick Hold-Geoffroy. I'm currently interning at Adobe again, focusing on editability and controllability of this lighting latent space.

Before joining Université Laval, I completed my MEng in Computer Technologies at South China University of Technology with Prof. Chuhua Xian, focusing on consistent depth estimation and BRDF estimation. During this period, I had the pleasure to intern as an Unreal Engine developer to develop my own mini-game at Alibaba Lingxi Interactive. I also worked as a rendering developer intern at Revobit, a start-up focusing on cloud rendering for the fashion industry. I got my BMgnt in E-Commerce from Xidian University.

CV / Email / GitHub / Google Scholar / LinkedIn

I expect to graduate in early/mid 2027 and am looking for full-time positions. Feel free to reach out!

hidden_bunny — Put and relight the bunny under the chair

Research

	UniLight: A Unified Representation for Lighting Zitian Zhang, Iliyan Georgiev, Michael Fischer, Yannick Hold-Geoffroy, Jean-François Lalonde, Valentin Deschaintre CVPR 2026 (highlight) arxiv / website / A joint latent space that unifies multiple lighting representations (environment maps, irradiance, spherical harmonics, text) for cross-modal retrieval, generation, and lighting control.
	Improving the color accuracy of lighting estimation models Zitian Zhang, Joshua Urban Davis, Jeanne Phuong Anh Vu, Jiangtao Kuang, Jean-François Lalonde CIC33 (Oral) arxiv / dataset / website / We propose a white-balanced HDR environment map dataset to evaluate the color robustness of lighting estimation models. We demonstrate that preprocessing inputs with a pre-trained white balance network significantly improves color accuracy without requiring retraining.
	SpotLight: Shadow-Guided Object Relighting via Diffusion Frédéric Fortier-Chouinard, Zitian Zhang, Louis-Etienne Messier, Mathieu Garon, Anand Bhattad, Jean-François Lalonde 3DV 2026 arxiv / code / dataset / website / supplementary / SpotLight enables object relighting by using shadow guidance to control lighting conditions without requiring additional training. Builds upon our previous work ZeroComp.
	ZeroComp: Zero-shot Object Compositing from Image Intrinsics via Diffusion Zitian Zhang, Frédéric Fortier-Chouinard, Mathieu Garon, Anand Bhattad, Jean-François Lalonde WACV 2025 (Oral) arxiv / code / dataset / poster / website / supplementary / Zero-shot 3D object compositing without paired training data. Easily extendable to 2D compositing and material editing.
	Delving into Multi-illumination Monocular Depth Estimation: A New Dataset and Method Yuan Liang, Zitian Zhang, Chuhua Xian, Shengfeng He TMM paper / code / dataset / An indoor RGB-D dataset with multi-illuminations, and a method for consistent depth estimation under changing lighting.