Xuhai Chen

I received my M.S. degree from Zhejiang University, China, in 2026, where I was advised by Prof. Yong Liu. Prior to that, I received my bachelor's degree from Tongji University, China.

In 2024, I interned at the State Key Laboratory of CAD&CG at Zhejiang University, where I worked on motion generation with Prof. Xiaowei Zhou and Prof. Sida Peng. In 2025, I worked as a remote research intern at the University of Pennsylvania, focusing on 3D reconstruction and generation with Prof. Lingjie Liu. Since July 2026, I have been collaborating remotely with Prof. Guandao Yang at UT Austin on 3D mesh generation and editing.

My research interests lie in 3D computer vision and graphics, with a particular focus on 3D generative modeling, reconstruction, and editing. I have also worked on image super-resolution and anomaly detection.

I enjoy exploring new experiences and challenges, which makes me feel happy and fulfilled. Life is meant to be experienced.

Email / Scholar / Github / CV

Awards

[2023.06]: CVPR 2023 workshop VAND Challenge: Winner in the Zero-shot Track, Honorable Mention in the Few-shot Track.

Papers

	Contact Matrix: Enhancing Dance Motion Synthesis with Precise Interaction Modeling Xuhai Chen, Zhi Cen, Huaijin Pi, Sida Peng, Xiaowei Zhou, Yong Liu CVPRF, 2026 paper / bibtex Generate realistic reactive motions for duet dance by modeling fine-grained leader–follower contacts with contact-guided diffusion.
	Better "CMOS" Produces Clearer Images: Learning Space-Variant Blur Estimation for Blind Image Super-Resolution Xuhai Chen, Jiangning Zhang, Chao Xu, Yabiao Wang, Chengjie Wang, Yong Liu CVPR, 2023 paper / github / bibtex Estimating space-variant blur degradation with the help of semantic information.
	A Zero-/Few-Shot Anomaly Classification and Segmentation Method for CVPR 2023 VAND Workshop Challenge Tracks 1&2: 1st Place on Zero-shot AD and 4th Place on Few-shot AD Xuhai Chen, Yue Han, Jiangning Zhang arXiv, 2023 paper / github / bibtex Technical report for the VAND challenge at the 2023 CVPR workshop.
	CLIP-AD: A Language-Guided Staged Dual-Path Model for Zero-shot Anomaly Detection Xuhai Chen, Jiangning Zhang, Guanzhong Tian, Haoyang He, Wuhao Zhang, Yabiao Wang, Chengjie Wang, Yong Liu arXiv, 2023 paper / github / bibtex Adapt the CLIP model for anomaly segmentation by merely fine-tuning a linear layer, and explain the text prompts design from a distributional perspective.
	Exploring Plain ViT Reconstruction for Multi-class Unsupervised Anomaly Detection Jiangning Zhang, Xuhai Chen, Yabiao Wang, Chengjie Wang, Yong Liu, Xiangtai Li, Ming-Hsuan Yang, Dacheng Tao arXiv, 2023 paper / github / bibtex Construct a reverse distillation architecture for multi-class anomaly detection using plain ViT.