ShengYun Peng

CS PhD (adversarial machine learning)

I research how to defend the machine learning attacks.
I have collaborated with researchers, developers, and scientists while working at Georgia Tech, Intel Lab, UCLA Design Automation Lab, Tongji International Joint Research Lab of Earthquake Engineering, Shanghai Jiao Tong University Advanced Avionics and Intelligent Information Lab, and Fudan University.

Education

Jan. 2022 — present
Ph.D. in Computer Science
Georgia Institute of Technology, Atlanta, GA
Advisor: Duen Horng (Polo) Chau
Jan. 2021 — present
M.S. in Computational Science & Engineering
Georgia Institute of Technology, Atlanta, GA
GPA: 4.00/4.00
Sept. 2015 — June 2020
B.Eng. in Civil Engineering
Tongji University, Shanghai, China
GPA: 92.37/100, Outstanding Graduates Award of Shanghai, 2020
July 2019 — Sept. 2019
Cross-disciplinary Scholars in Science and Technology (CSST) Program
University of California, Los Angeles, Los Angeles, CA
GPA: 4.00/4.00

Academic Research Experience

May 2021 — present
Georgia Institute of Technology, Atlanta, GA
Graduate Research Assistant, Polo Club of Data Science
Mentor: Duen Horng (Polo) Chau
Working on the DARPA Guaranteeing AI Robustness Against Deception (GARD) project
July 2019 — Dec. 2019
University of California, Los Angeles, Los Angeles, CA
Undergraduate Research Assistant, Design Automation Lab
Mentor: Lei He
We proposed a novel anchor free tracker, which represented an object by a single point at bounding box center, and other properties such as object size and local offsets were regressed directly from image features at the center location. I simulated operation fusion, network slicing, and data quantization for tracking network in PyTorch. We also worked on the deployment of trackers on FPGA boards with the assistance of an RTL based hand-coded FPGA overlay domain-specific processor unit (OPU) with software programmability and fast compilation time, targeting at general CNN accelerations.
July 2018 — Jan. 2020
Shanghai Jiao Tong University, Shanghai, China
Undergraduate Research Assistant, Advanced Avionics and Intelligent Information Lab
Mentor: Gang Xiao
We developed both pixel-level and feature-level visible and infrared fusion methods based on Siamese networks, which solved tracking challenges like partial occlusion, low illumination. The tracker ranked first on large-scale RGBT234 dataset.
Oct. 2017 — Aug. 2020
Tongji University, Shanghai, China
Undergraduate Research Assistant, State Key Laboratory of Disaster Reduction in Civil Engineering
Mentor: Ying Zhou
I proposed a novel and complete framework which obtained the natural frequency and modes of vibration with a non-contact measurement method based on optical flow, correlation filter and Siamese network with single UAV. We measured the displacements of a five-storey structure and the errors of first order and the second order natural frequency were 0.09% and 1.01%, respectively. We also verified the accuracy through the first shaking table test of a full scale low-damage concrete wall building in the world at International Joint Research Laboratory of Earthquake Engineering, and the mean squared error of UAV camera and linear variable differential transformers was below 4%
Sept. 2020 — Dec. 2020
Fudan University, Shanghai, China
Undergraduate Research Assistant, State Key Laboratory of Disaster Reduction in Civil Engineering
Mentor: Weidong Yang
We developed a vision monitoring software to check the layout in cockpit, and automatically verify the correctness of parameters in airspeed indicator, vertical speed indicator, and attitude indicator.
June 2017 — Aug. 2017
Chinese Academy of Sciences, Beijing, China
Undergraduate Research Assistant, Institute of Computing Technology
Mentor: Chao Liu
We built a Simulink model to simulate missile vertical launch and six-freedom system for aircraft flight control. We also illustrated the changes in pitch angle and posture adjustment with the involvement of PID control system when disturbed by atmosphere, aiming to realize assistive and full-automatic flight path control.
Apr. 2017 — Apr. 2018
Tongji University, Shanghai, China
Undergraduate Research Assistant, School of Materials Science and Engineering
Mentor: Qianrong Yang
We finished the mechanical design and manufacture of the 3D printer with automatic pulp feeding systems, assisted by BIM to visualize construction process and cost. I also performed sample rupture strength and tensile strength tests to modify the working path in order to achieve the desired physical properties.

Honors and Awards

2017 — 2018
National Scholarship
Stipend awarded to top 2% undergraduates in China
2017 — 2018
Tongji University Excellent Student
For top 5% undergraduates in Tongji University
2016 — 2017
First Prize of Tongji University Scholarship of Excellence
For outstanding undergraduates in Tongji University
2016 — 2017
Excellent Worker in the ASCE-Tongji-ISG of the Instructional Innovation
For my work at ASCE-Tongji-ISG club
2017
National Undergraduate Innovative Programs Certificate
For our research in “3D Printing of Concrete Structures and Sample Strength Tests”
2017
3rd Prize in Contemporary Undergraduate Mathematical Contest in Modeling
We worked on the second problem, which was creating a price model for a lately photo service
2016
1st Prize in Mathematics Competition of Chinese College Students
The competition tests calculus and linear algebra

Publications

Selected: Latest & Greatest

A novel DNN tracking algorithm for structural system identification
ShengYun Peng, LingFeng Yan, Bin He, Ying Zhou
Smart Structures and Systems. 2021.
Project PDF BibTeX DOI
Accurate Anchor Free Tracking
ShengYun Peng, YunXuan Yu, Kun Wang, Lei He
arXiv:2006.07560. 2020.
Project PDF BibTeX

Journal

J4
A novel DNN tracking algorithm for structural system identification
ShengYun Peng, LingFeng Yan, Bin He, Ying Zhou
Smart Structures and Systems. 2021.
Project PDF BibTeX DOI
J3
DSiamMFT: An RGB-T fusion tracking method via dynamic Siamese networks using multi-layer feature fusion
Xingchen Zhang, Ping Ye, ShengYun Peng, Jun Liu, Gang Xiao
Signal Processing, Image Communication. 2020.
Project PDF BibTeX DOI
J2
Anti-occlusion object tracking based on correlation filter
Jun Liu, Gang Xiao, Xingchen Zhang, Ping Ye, Xingzhong Xiong, ShengYun Peng
Signal, Image and Video Processing. 2019.
Project PDF BibTeX DOI
J1
SiamFT: An RGB-Infrared Fusion Tracking Method via Fully Convolutional Siamese Networks
Xingchen Zhang, Ping Ye, ShengYun Peng, Jun Liu, Ke Gong, Gang Xiao
IEEE Access. 2019.
Project PDF BibTeX DOI

Conference

C1
Object Fusion Tracking Based on Visible and Infrared Images Using Fully Convolutional Siamese Networks
Xingchen Zhang, Ping Ye, Dan Qiao, Junhao Zhao, ShengYun Peng, Gang Xiao
22th International Conference on Information Fusion (FUSION). 2019.
Project PDF BibTeX DOI

Miscellaneous

M1
Accurate Anchor Free Tracking
ShengYun Peng, YunXuan Yu, Kun Wang, Lei He
arXiv:2006.07560. 2020.
Project PDF BibTeX

Service

Reviewer
Pattern Recognition (Pattern Recognit.) 2020
IEEE Access 2020
Member
2016 — 2017
American Society of Civil Engineers - Tongji (ASCE)

References

Dr. Polo Chau, Associate Professor
School of Computational Science and Engineering
Georgia Institute of Technology
Dr. Lei He, Professor
School of Electrical and Computer Engineering
University of California, Los Angeles
Dr. Ying Zhou, Professor
College of Civil Engineering
Tongji University