Zhanghan Ke

Ph.D. Student @ CityU


I am Zhanghan Ke (柯章翰), a Ph.D. student in the Department of Computer Science, City University of Hong Kong (CityU), under the supervision of Prof. Rynson W.H. Lau. Before that, I obtained my B.Eng. from Software College, Northeastern Universiy (NEU) in 2018.

My current research interests are in the areas of semi-/weak-/un-supervised learning, transfer learning, and their applications in computer vision & graphics.

lychee Google Scholar / GitHub / Linked In

Research Experiences

Mar. 2018 - Apr. 2021

Research Intern (with Excellence Award and Future Star Award) in AI Imaging Group, SenseTime Research, supervised by Dr. Qiong Yan.

Nov. 2016 - Aug. 2017

Research Intern (with Excellence Award) in Internet Graphics Group, Microsoft Research Asia (MSRA), supervised by Dr. Xin Tong and Dr. Yuxiao Guo.

Selected Publications

Is a Green Screen Really Necessary for Real-Time Portrait Matting?

Zhanghan Ke, Kaican Li, Yurou Zhou, Qiuhua Wu, Xiangyu Mao, Qiong Yan, Rynson W.H. Lau

Arxiv Preprint, 2020

paper / code / supp video

Towards Geometry Guided Neural Relighting with Flash Photography

Di Qiu, Jin Zeng, Zhanghan Ke, Wenxiu Sun, Chengxi Yang

International Conference on 3D Vision (3DV), 2020


Guided Collaborative Training for Pixel-wise Semi-Supervised Learning

Zhanghan Ke, Di Qiu, Kaican Li, Qiong Yan, Rynson W.H. Lau

European Conference on Computer Vision (ECCV), 2020

paper / code / video

Dual Student: Breaking the Limits of the Teacher in Semi-Supervised Learning

Zhanghan Ke, Daoye Wang, Qiong Yan, Jimmy Ren, Rynson W.H. Lau

International Conference on Computer Vision (ICCV), 2019

paper / code / poster

Open Source Projects


GitHub Repository

PixelSSL is a PyTorch-based semi-supervised learning (SSL) codebase for pixel-wise (Pixel) vision tasks. The purpose of this project is to promote the research and application of semi-supervised learning on pixel-wise vision tasks. It provides (1) an interface for implementing new semi-supervised algorithms and (2) a template for encapsulating diverse computer vision tasks. As a result, the SSL algorithms integrated in PixelSSL are compatible with all task codes inherited from the given template.

More About Me

Teaching Assistant

2019/20 Semester A - CS4185 Multimedia Techs & Appls

2019/20 Semester B - CS4188/CS5188 Virtual Reality Tech & Apns

A Little Guy


Hey! Look here! I am a cute kitty who lives with Zhanghan. (ノ>ω<)ノ

But I will not tell you that my name is hidden somewhere on this page. ( • ̀ω•́ )