Publications

The first ImageNet classifier that can survive strong white-box adversarial attacks.
Conference on Computer Vision and Pattern Recognition (CVPR), 2019

A batch-independent alternative to Batch Normalization.
European Conference on Computer Vision (ECCV), 2018 (Oral),
Best Paper Honorable Mention

Workshop in International Conference on Learning Representations (ICLR), 2018

Conference on Neural Information Processing Systems (NeurIPS), 2017 (Oral)

Award-winning solution for training AI to play FPS games.
International Conference on Learning Representations (ICLR), 2017

Comprehensive results on low bitwidth neural network.
Tech Report, 2016

Open Source Projects

More at github.

A neural net training interface on TensorFlow, with focus on speed + flexibility.

Black-box adversarial attacks on AWS/Azure’s public face recognition APIs, the first successful attack of its kind.

The best-performing open source TensorFlow reimplementation of Faster/Mask/Cascade R-CNN.

Panorama stitching written from scratch in C++.

Export encrypted wechat message history from Android phones by reverse-engineering.

Photorealistic rendering with Monte Carlo path tracing.

Work Experience

 
 
 
 
 
March 2017 – Present
Menlo Park

Research Engineer

Facebook AI Research

Research & systems in computer vision & reinforcement learning.
 
 
 
 
 
January 2016 – April 2016
Pittsburgh

Research Assistant

Oculus Research

Generative modeling.
 
 
 
 
 
October 2014 – August 2015
Beijing

Researcher

Megvii (aka Face++)

Research & products in computer vision.
Neural network training framework.
 
 
 
 
 
April 2014 – July 2014
Beijing

Software Engineering Intern

Google

SfM (structure from motion) from videos.

Awards

  • Winner of defense track in Competition on Adversarial Attacks and Defenses (CAAD) 2018.

    We trained an ImageNet classifier with state-of-the-art robustness against adversarial attacks. Competition

  • Winner of CTF in Competition on Adversarial Attacks and Defenses (CAAD) CTF 2018.

    We performed successful adversarial attacks / defenses against other teams during the on-site competition. Competition Media (Chinese)

  • Best Paper Honorable Mention award in ECCV 2018. Announcement

  • Google Open Source Peer Bonus in 2017. Announcement

  • Winner of VizDoom AI Competition in CIG 2016.

    Our Doom bot, “F1”, beat all other competitors in a death match by a large margin. Competition Media Video

  • Champion of Student Cluster Competition in both ISC 2015 and ASC 2015.

    Optimize software and hardware for low-power high-performance computation. Media Media

  • Finalist in SIGMOD Programming Contest in SIGMOD 2014.

    Design and implement a large social network database to support efficient queries. Competition Slides Poster

  • Capture the Flags (CTF, a form of security competition): 11th in SECCON CTF 2014, 8th in Codegate CTF 2015, 5th in DEFCON CTF 2015 as a team member of blue-lotus.