Shen Zhuoran (Zhuoran is the first name) is an AI Resident at Google Research. He holds a BEng in Computer Science from The University of Hong Kong and has formerly been Research Interns at Tencent and SenseTime. He currently focuses on the the attention mechanism, including the application of efficient attention in various domains and fully-attentional visual modeling. He also has interests in stereo vision tasks and natural language processing.

Education

The University of Hong Kong, Hong Kong

  • Sep. 2015 - Jun. 2019.
  • Bachelor of Engineering in Computer Science.
  • GPA: 3.85/4.30. Standing: 1/111. Major GPA: 3.96/4.30.

University of California, Davis, Davis, CA, Unites States

  • Sep. 2017 - Dec. 2017.
  • Bachelor’s Reciprocity Student in Computer Science.
  • GPA: 4.00/4.00.

Work Experience

Google, Seattle, WA, United States

  • Oct. 2019 - Present.
  • AI Resident, Cerebra Team, Google AI
  • Working on foundations of fully-attentional visual modeling. Prepared a submission to NeurIPS 2020.

Tencent, Shenzhen, Guangdong, China

  • Jul. 2019 - Sep. 2019.
  • Research Intern, Applied Research Center, Platform and Content Group
  • Designed and validated the global context module. Details in Research Experience

SenseTime, Hong Kong

  • Jun. 2017 - Jun. 2019.
  • Research Intern, Intelligent Perception and Services Team, Smart City Group
  • Conducted academic research projects listed in Research Experience.
  • Worked on the inception of an autonomous driving project, including task identification, data scheme de-sign, coordination of collection and labeling, and algorithmic design and validation.

Awards

  • Dean’s Honours List 2017-2018, Faculty of Engineering, The University of Hong Kong
  • Dean’s Honours List 2016-2017, Faculty of Engineering, The University of Hong Kong
  • Dean’s Honours List 2015-2016, Faculty of Engineering, The University of Hong Kong
  • Dean’s Honor List, Fall Quarter 2017, College of Letters and Science, University of California, Davis
  • YC Cheng Engineering Scholarship, 2017, Faculty of Engineering, The University of Hong Kong

Programming Contests

  • First Runner-up, ACM-HK Programming Contest 2017
  • Second Runner-up, ACM-ICPC Hong Kong PolyU International Invitational 2017
  • Second Runner-up, hackUST 2017 Radica Challenge
  • First Prize, National Olympiad of Informatics in Provinces (China) 2014

Research Experience

Global Context Module, Tencent

  • Jul. 2019 - Sep. 2019.
  • Supervised by Dr. Shan Ying, Director of Applied Research Center, Platform and Content Group, Tencent.
  • Proposed the global context module, which uses efficient attention to achieve linear complexities in spatial size and constant complexities in temporal duration for deep video memory.
  • Built the first real-time video object segmenter that has state-of-the-art accuracy (86.6, J&F @ 25 FPS, DAVIS 2016).
  • To present a first-author paper at ECCV 2020.

Efficient Attention, Industry Research Experience

  • Sep. 2018 - Jun. 2019.
  • Supervised by Dr. Yi Shuai, Research Director, SenseTime.
  • In collaboration with Dr. Li Hongsheng, Assistant Professor, Multimedia Laboratory, Chinese University of Hong Kong.
  • Proposed efficient attention, which reduced the memory and computational complexities of the attention mechanism from quadratic to linear and is applicable to computer vision, natural language processing (NLP), and speech analysis.
  • Achieved new states-of-the-art on object detection (41.8→43.1, AP, COCO 2017) and stereo depth estimation (1.09→0.477, EPE, Scene Flow) and significant improvement on instance segmentation (36.6→37.9, AP, COCO 2017) and image classification (93.0%→93.7%, top-5 accuracy, ImageNet).
  • Submitted a first-author paper to WACV 2021.

Heterogeneous Graph Neural Network, Industry Research Experience

  • Mar. 2019 - Jun. 2019.
  • Supervised by Zhao Haiyu, Senior Research Scientist, SenseTime.
  • Proposed heterogeneous graph neural network, which uses instance-instance attention and pixel-instance attention modules to enhance contextual information for visual relationship detection.
  • Set new states-of-the-art on visual relationship detection (52.0→52.9, role AP, V-COCO; 33.91→38.79, recall@100&k=70, VRD).

Visual Embedding of Chinese, Bachelor’s Final-Year Project

  • Sep. 2018 - Apr. 2019.
  • Supervised by Dr. Kwan-Yee Kenneth Wong, Associate Professor, Computer Vision Group, The University of Hong Kong.
  • Designed OceanText, a novel character embedding algorithm for Chinese that extracts a semantic embedding from the image of a Chinese character with a convolutional neural network.
  • Developed a PyTorch embedding library. Reduced single-GPU training time from 82 days to 28.1 hours compared to existing open-source implementations.
  • Set a new state-of-the-art for word similarity estimation from character embeddings (15.6%→37.6 Spearman’s ρ, WordSim-297 Chinese).

Diabetic Retinopathy Analysis, Industry Research Experience

  • Sep. 2018 – Oct. 2018.
  • Supervised by Zhao Haiyu, Senior Research Scientist, SenseTime
  • Experimented with different hyperparameters and network architectures.
  • Improved the accuracy on the IDRiD dataset from 40.5% to 59.6%.

Stereo Depth Estimation, Industry Research Experience

  • May 2018 - Sep. 2018.
  • Supervised by Dr. Yi Shuai, Research Director, SenseTime.
  • Developed a stereo depth estimator based on PSMNet with improved training procedures.
  • Achieved 2x reduction in error rate (EPE) over the previous state-of-the-art on the Scene Flow dataset.

Teaching Experience

Software Engineering, Teaching Assistant

  • Jan. 2019 - May 2019.
  • Assisted George Mitcheson, Guest Lecturer, Department of Computer Science, The University of Hong Kong.
  • Developed a Django server as the external HR server for student projects and deployed it to Heroku.
  • Answered questions from and held consultations with students on Git, the Unified Modeling Language, and software design and engineering principles.

Personal Projects

BeautyNet

  • May 2018 - Oct. 2019.
  • Owner and primary contributor.
  • Developed a PyTorch project template. Applied deduplication, modularization, and a consistent code style to improve maintainability, testability, and analyzability.
  • Became the 2nd most popular PyTorch template on GitHub, got 180+ stars, and trended for 3 days.

Preprints

Skills

  • Programming: Python, C, C++, Java, Shell script, Markdown, LaTeX
  • Machine Learning: PyTorch, TensorFlow, Keras, Sonnet, Caffe, CUDA, NumPy, OpenCV
  • Technologies: Git, Vim, Slurm, Django, Jekyll, Piper, Blaze, Gin
  • Languages: Mandarin Chinese (native), English (working proficiency, 116 in TOEFL)