Zijian Hu

Research Engineer

ByteDance

Biography

Zijian Hu is a research engineer at ByteDance Intelligent Creation Lab. He is interested in building robust machine perception systems that can operate in noisy environments and can be trained and adapted with minimal supervision.

Download my CV and resumé (last updated on May 21, 2021) for more details.

Interests

Computer Vision
Semi-Supervised Learning
Multimodal Machine Learning
Low-Shot Learning
Robotics
Human-Robot Interaction
Robot Learning

Education

MSc in Computer Science, 2020

University of Southern California
BSc in Computer Science, 2020

University of Southern California

Skills

Machine Learning

PyTorch, TensorFlow, Keras, OpenCV

Robotics

ROS, OpenCV, V-Rep

Mathematics

Statistical Learning, Probability, Calculus

Programming Languages

Python, C/C++, Java, JavaScript (ES6), MATLAB

Web Development

Node.js, Java EE, Angular

Computer System & Hardware

X86 Assembly/GAS, MIPS, Verilog, Arduino

Experience

Research Engineer, Computer Vision

ByteDance

Oct 2021 – Present California

Worked on the following topics at ByteDance Intelligent Creation Lab:
- Computer Vision
- Video Content Understanding
- Semi-Supervised Learning
- Self-supervised Learning
- Object Detection

Research Staff

IRIS Computer Vision Lab

Sep 2020 – Oct 2021 California

Advised by Professor Ram Nevatia
Research Area:
- Computer Vision
- Semi-Supervised Learning
- Low-Shot Learning
- Unsupervised Learning
Projects:
- Learning with Less Labeling (LwLL)

Research Assistant

IRIS Computer Vision Lab

Jun 2019 – May 2020 California

Advised by Professor Ram Nevatia
Research Area:
- Computer Vision
- Low-Shot Learning
- Meta Learning
Projects:
- Learning with Less Labeling (LwLL)

Research Assistant

ICAROS Lab

Apr 2019 – May 2020 California

Advised by Professor Stefanos Nikolaidis
Research Area:
- Robotics
- Human-Robot Interaction
- Computer Vision
Projects:
- Resource Distribution in Human-Robot Teams

Research Assistant

Interaction Lab

May 2018 – May 2020 California

Advised by Professor Maja Matarić
Research Area:
- Human-Robot Interaction
- Multimodal Machine Learning
- Computer Vision
Projects:

Full Stack Engineer Intern

4Paradigm

Jun 2017 – Aug 2017 Beijing, China

Developed an office automation software using Node.js (with koa) and Angular which:

Allows authorized users to manage stored employee information
Allows authorized users to manage stored department information
Has configurable user access control
Support user registration and authentication. Submitted user profiles are stored and waited to be approved by authorized personnel

Android Engineer Intern

Nox Limited

Jun 2016 – Jul 2016 Beijing, China

Developed a tutorial Android application for NoxPlayer, an android emulator for PC

Featured Publications

Zijian Hu, Zhengyu Yang, Xuefeng Hu, Ram Nevatia

June 2021 2021 Conference on Computer Vision and Pattern Recognition (CVPR)

SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification

We propose a novel algorithm that focuses on the less studied relationship between the unlabeled data. Our algorithm achieves state-of-the-art performance on standard benchmarks

Chris Birmingham, Zijian Hu, Kartik Mahajan, Eli Reber, Maja J. Matarić

May 2020 2020 International Conference on Robotics and Automation (ICRA)

Can I Trust You? A User Study of Robot Mediation of a Support Group

We designed and evaluated a novel framework for robot mediation of a support group

Recent Publications

Quickly discover relevant content by filtering publications.

Naomi T. Fitter, Youngseok Joung, Marton Demeter, Zijian Hu, Maja J. Matarić (2019). Design and Evaluation of Expressive Turn-Taking Hardware for a Telepresence Robot. 2019 IEEE International Symposium on Robot and Human Interactive Communication (Ro-Man).

PDF Cite Project Video DOI

Naomi T. Fitter, Youngseok Joung, Zijian Hu, Marton Demeter, Maja J. Matarić (2019). User Interface Tradeoffs for Remote Deictic Gesturing. 2019 IEEE International Symposium on Robot and Human Interactive Communication (Ro-Man).

PDF Cite Project DOI