Yuanze Lin

I am a PhD student in the Computer Science Department at the University of Oxford, where I work with Prof. Ronald Clark and Prof. Philip Torr. I also collaborate with Dr. João F. Henriques from the VGG group. My research focuses on diffusion and vision-language models.

Before going to Oxford, I had a great time at Microsoft Redmond, MSR Asia, CCVL @ Johns Hopkins University, Alibaba, etc. I appreciate collaborating with distinguished professors and researchers from these institutions.

My research interest lies in machine learning and its applications, especially:

• Multimodal large language models (MLLMs)

• Image/Video/3D generation and editing based on diffusion models

• The applicability of large language models (LLMs)

yuanze.lin [at] cs.ox.ac.uk / Google Scholar / GitHub / LinkedIn

News

[03/2025] Released the code of Olympus (CVPR 2025) for vision tasks.

[02/2025] Olympus accepted to CVPR 2025!

[12/2024] We presented Olympus to solve over 20 different computer vision tasks.

[08/2024] Released the code of Learnable Regions (CVPR2024) for image editing.

[07/2024] Rethinking Visual Prompting for MLLMs with External Knowledge was presented.

[03/2024] Check out DreamPolisher for high-quality text-to-3D generation!

[02/2024] Started a research internship at GenAI @ Microsoft.

[02/2024] Text-Driven Image Editing via Learnable Regions accepted to CVPR 2024.

[10/2023] Started my PhD journey at CS @ University of Oxford.

[07/2023] SMAUG accepted to ICCV 2023.

[09/2022] REVIVE accepted to NeurIPS 2022.

[03/2022] Pseudo-Q and AdaFocus V2 accepted to CVPR 2022.

[07/2021] MCN accepted to ICCV 2021.

[06/2021] EVA-GCN accepted to CVPR 2021 AMFG Workshop and won Best Paper Award!

Selected Publications

Papers are sorted by recency, * denotes equal contribution.

	Olympus: A Universal Task Router for Computer Vision Tasks Yuanze Lin, Yunsheng Li, Dongdong Chen, Weijian Xu, Ronald Clark, Philip H.S. Torr CVPR, 2025 ArXiv / Project Page / Code / BibTeX We introduce Olympus, a new approach that transforms Multimodal Large Language Models (MLLMs) into a unified framework capable of handling a wide array of computer vision tasks.
	Text-Driven Image Editing via Learnable Regions Yuanze Lin, Yi-Wen Chen, Yi-Hsuan Tsai, Lu Jiang, Ming-Hsuan Yang CVPR, 2024 ArXiv / Project Page / Video / Code / BibTeX Introduce a region-based editing network that is trained to generate editing regions utilizing a text-driven editing loss with CLIP guidance, our method can edit the given images based on freely provided language descriptions.
	SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-training Yuanze Lin, Chen Wei, Huiyu Wang, Alan Yuille, Cihang Xie ICCV, 2023 ArXiv / Poster / Slides / BibTeX Propose an efficient video-language pre-training framework, which enjoys both competitive performances on text-to-video retrieval and video question answering tasks, and much less pre-training costs by 1.9X or more.
	REVIVE: Regional Visual Representation Matters in Knowledge-Based Visual Question Answering Yuanze Lin, Yujia Xie, Dongdong Chen, Yichong Xu, Chenguang Zhu, Lu Yuan NeurIPS, 2022 ArXiv / Poster / Supplementary Material / OpenReview / Code / BibTeX Propose a new knowledge-based VQA method REVIVE, which utilizes the explicit information of object regions not only in the knowledge retrieval stage but also in the answering model. It achieves a new state-of-the-art performance on OK-VQA dataset.
	Pseudo-Q: Generating Pseudo Language Queries for Visual Grounding Haojun Jiang, Yuanze Lin, Dongchen Han, Shiji Song, Gao Huang CVPR, 2022 ArXiv / Poster / Code / BibTeX Present Pseudo-Q to automatically generate pseudo language queries for supervised training, which achieves superior or comparable performance compared to existing weakly-supervised visual grounding methods on five datasets.
	AdaFocus V2: End-to-End Training of Spatial Dynamic Networks for Video Recognition Yulin Wang, Yang Yue, Yuanze Lin, Haojun Jiang, Zihang Lai, Victor Kulikov, Nikita Orlov, Humphrey Shi, Gao Huang CVPR, 2022 ArXiv / Code / BibTeX Reformulate AdaFocus as a simple one-stage algorithm by introducing a differentiable interpolation-based patch selection operation and further present an improved training scheme. Extensive experiments on six benchmark datasets demonstrate its effectiveness.
	Self-supervised video representation learning with meta-contrastive network Yuanze Lin, Xun Guo, Yan Lu ICCV, 2021 ArXiv / Poster / BibTeX Propose a Meta-Contrastive Network (MCN), which combines contrastive learning and meta learning for pre-training. For video action recognition and video retrieval tasks, MCN outperforms state-of-the-art approaches on UCF101 and HMDB51 datasets.
	EVA-GCN: Head Pose Estimation Based on Graph Convolutional Networks Miao Xin, Shentong Mo, Yuanze Lin CVPR AMFG Workshop, 2021 (Best Paper Award) Paper / Code / BibTeX Construct a landmark-connection graph, and propose to leverage the Graph Convolutional Networks (GCN) to model the complex nonlinear mappings between the graph typologies and the head pose angles.

Experiences

	GenAI @ Microsoft Redmond Researcher Intern, Feb 2024 - Nov 2024 hosted by Dr. Dongdong Chen, working on large multimodal models.
	Vision and Learning Lab @ UC Merced Visiting Student, May 2022 - Nov 2023 hosted by Prof. Ming-Hsuan Yang, working on text-driven image editing.
	Alibaba Group Senior Algorithm Engineer, Feb 2023 - Aug 2023 Working on vision-language pre-training, fine-tuning, and the applicability of large language models (LLMs).
	CCVL @ Johns Hopkins University Research Assistant, May 2022 - Feb 2023 with Prof. Cihang Xie and Prof. Alan Yuille, working on vision-language pre-training based on MAE.
	Microsoft Redmond Researcher Intern, Feb 2022 - June 2022 with Dr. Yujia Xie, Dr. Dongdong Chen and Dr. Yichong Xu, working on knowledge-based VQA.
	Microsoft Research Asia Researcher Intern, Dec 2020 - Sep 2021 with Dr. Xun Guo and Dr. Yan Lu, working on self-supervised learning and transformers for video tasks.
	Tencent AI Lab Researcher Intern, Sep 2020 - Dec 2020 with Dr. Haozhi Huang, working on text-driven editing of videos based on meta learning.