Welcome to Jianyu Wang's Homepage!

I am Jianyu Wang, currently a Software Engineer at Waymo, working on object detection for cars.

Before joining Waymo, I was a Research Scientist at Baidu Research USA. I obtained my Ph.D. degree from the Statistics Department of UCLA in March 2017. During my Ph.D., I was very fortunate to work with Prof. Alan Yuille in computer vision and deep learning. I obtained my master degree from the Electrical Engineering Department of UCLA in June 2013, where I majored in wireless communication and optimization. Before that, I did my undergraduate in the Electronic Engineering Department of Tsinghua University from 2007 to 2011. I did summer internship twice at Google, Mountain View, in 2015 and 2016.

Email: wjyouch[AT]gmail[DOT]com

[resume] [google scholar]

Research Interests

My research interests generally lie in computer vision and deep learning. Specifically, I take interests in:

  • Adversarial Attack and Defense in Machine Learning

  • Part and Composition, Object Recognition

  • Deep Learning Visualization and Understanding

  • Incremental / Continue learning



  • Joint Adversarial Training: Incorporating both Spatial and Pixel Attacks. arXiv, 2019.

Haichao Zhang, Jianyu Wang


  • Adversarial Attacks and Defences Competition. arXiv, 2018.

Alexey Kurakin, Ian Goodfellow, Samy Bengio, Yinpeng Dong, Fangzhou Liao, Ming Liang, Tianyu Pang, Jun Zhu, Xiaolin Hu, Cihang Xie, Jianyu Wang, Zhishuai Zhang, Zhou Ren, Alan Yuille, Sangxia Huang, Yao Zhao, Yuzhe Zhao, Zhonglin Han, Junjiajia Long, Yerkebulan Berdibekov, Takuya Akiba, Seiya Tokui, Motoki Abe


  • Unsupervised Learning of Object Semantic Parts from Internal States of CNNs by Population Encoding. arXiv, 2016.

Jianyu Wang, Zhishuai Zhang, Cihang Xie, Vittal Premachandran, Alan Yuille



  • Enhancing Cross-task Black-box Transferability of Adversarial Examples with Dispersion Reduction. CVPR, 2020.

Yantao Lu, Yunhan Jia, Jianyu Wang, Bai Li, Weiheng Chai, Lawrence Carin, Senem Velipasalar


  • Compositional Generalization for Premitive Substitutions. EMNLP-IJCNLP, 2019.

Yuanpeng Li, Liang Zhao, Jianyu Wang and Joel Hestness


  • Defense Against Adversarial Attacks Using Feature Scattering-based Adversarial Training. NeurIPS, 2019.

Haichao Zhang, Jianyu Wang

[paper, code]

  • Bilateral Adversarial Training: Towards Fast Training of More Robust Models Against Adversarial Attacks. ICCV, 2019.

Jianyu Wang, Haichao Zhang

[paper, code]

  • Towards Adversarially Robust Object Detection. ICCV, 2019.

Haichao Zhang, Jianyu Wang

[paper, code soon]

  • Improving Transferability of Adversarial Examples with Input Diversity. CVPR, 2019.

Cihang Xie, Zhishuai Zhang, Yuyin Zhou, Song Bai, Jianyu Wang, Zhou Ren, Alan Yuille


  • DeepVoting: An Explainable Framework for Semantic Part Detection under Partial Occlusion. CVPR, 2018.

Zhishual Zhang*, Cihang Xie*, Jianyu Wang*, Lingxi Xie, Alan Yuille (*equal contribution)


  • Mitigating adversarial effects through randomization. ICLR, 2018.

Cihang Xie, Jianyu Wang, Zhishual Zhang, Zhou Ren, Alan Yuille

Ranked No.2 in the NIPS 2017 Challenge: Defense against Adversarial Attack

[paper] [code]

  • Adversarial Examples for Semantic Segmentation and Object Detection. ICCV, 2017.

Cihang Xie*, Jianyu Wang*, Zhishuai Zhang*, Yuyin Zhou, Lingxi Xie, Alan Yuille (*equal contribution)

[paper] [code]

  • Detecting Semantic Parts on Partially Occluded Rigid Objects. BMVC, 2017.

Jianyu Wang*, Cihang Xie*, Zhishuai Zhang*, Lingxi Xie, Jun Zhu, Alan Yuille (*equal contribution)

[paper] [VehicleSemanticPart dataset for semantic part detection], [VehicleOcclusion dataset for part detection under occlusion]

  • Semantic Part Segmentation using Compositional Model combining Shape and Appearance. CVPR, 2015.

Jianyu Wang, Alan Yuille

[paper], [HorseCow dataset for shape analysis]


  • Visual Concepts and Compositional Voting. Annals of Mathematical Sciences and Applications, 2018.

Jianyu Wang, Zhishual Zhang, Cihang Xie, Yuyin Zhou, Vittal Premachandran, Jun Zhu, Lingxi Xie, Alan Yuille

[paper] This journal paper combines two conference papers into a complete framework.