Zhenmei ShiPh.D.
Computer Science
|
|
I am currently a second-year Ph.D. student majoring in Computer Science at the University of Wisconsin-Madison advised by Prof. Yingyu Liang. I obtained my B.S. degree, majoring in Computer Science and Pure Mathematics Advanced, from the Hong Kong University of Science and Technology in 2019.
I am fortunate to be a returning intern in the Google YouTube Ads Machine Learning team, Mountain View, in summer 2021. I had a wonderful experience in the Google Pixel Camera team, in the summer of 2020. Previously, I was an intern in the Base Model group of Megvii (Face++), Beijing, in the summer of 2019. Also, I had two internships at Tencent YouTu group, Shenzhen, in the winter of 2019 and the winter of 2018.
My research interest includes Machine Learning Theory and Computer Vision.
On Feature Learning in Neural Networks: Emergence from Inputs and Advantage over Fixed Features
Zhenmei Shi, Junyi Wei , Yingyu Liang (under review) We provide new insights on interesting phenomena of the loss landscape and training dynamics, including a wedge-like structure of the low-loss solutions, and the different roles of input structure for shallow and deep networks. |
|
Deep Online Fused Video Stabilization
Zhenmei Shi, Fuhao Shi, Wei-Sheng Lai, Chia-Kai Liang, Yingyu Liang (under review) [ Project ] [ Paper ] We present a deep neural network (DNN) that uses both sensor data (gyroscope) and image content (optical flow) to stabilize videos through unsupervised learning. |
|
Structured Feature Learning for End-to-End Continuous Sign Language Recognition Zhaoyang Yang*, Zhenmei Shi*, Xiaoyong Shen, Yu-Wing Tai arXiv, 2019 [ News ] [ Paper ] Our SF-Net which extracts features in a structured manner and gradually encodes information at the frame, the gloss and the sentence level into the feature representation. |
|
Dual Augmented Memory Network for Unsupervised Video Object Tracking
Zhenmei Shi*, Haoyang Fang*, Yu-Wing Tai, Chi Keung Tang arXiv, 2019 [ Project ] [ Paper ] Our Dual Augmented Memory Network is unique in remembering both target and background, and using an attention LSTM memory to guide the focus on memorized features. |
Velocity Vector Preserving Trajectory Simplification
Guanzhi Wang*, Zhenmei Shi*, Cheng Long, Ya Gao, Raymond Wong Technical Report, 2018 [ Code ] [ Paper ] We designed an algorithm to simplify a trajectory such that the number of points is minimized under the constraint that the velocity-based error does not exceed a given tolerance. |
|
Deep Colorization, 2018.
Advisor: Xin Tao and Prof. Yu-Wing Tai [ News ] |
|
High-Performance Traffic Assignment Based on Variational Inequality, 2017.
Advisor: Dr. Cheng Liu and Prof. Kwai L. Wong [ Poster ] |
Research Assistant
University of Wisconsin-Madison 2019 - Now | Prof. Yingyu Liang |
|
Software Engineering Intern
Google in Mountain View, CA Summer 2021 | Myra Nam Summer 2020 | Fuhao Shi |
|
Research Intern
Megvii (Face++) in Beijing Summer 2019 | Xiangyu Zhang |
|
Research Intern
Tencent YouTu in Shenzhen Winter 2019 | Zhaoyang Yang and Prof. Yu-Wing Tai Winter 2018 | Xin Tao and Prof. Yu-Wing Tai |
|
Research Assistant
Hong Kong University of Science and Technology 2018 - 2019 | Prof. Chi Keung Tang 2017 - 2018 | Prof. Raymond Wong Summer 2016 | Prof. Ji-Shan Hu |
|
Research Intern
Oak Ridge National Laboratory in the USA Summer 2017 | Dr. Cheng Liu and Prof. Kwai L. Wong |
Conference Reviewer at ICCV 2021 | |
Conference Reviewer at CVPR 2021 | |
Conference Reviewer at ECCV 2020 |