Machine Learning

NA ZHANG

Redmond, WA ***** j +1-860-***-**** j **********@*****.***

https://nz0001na.github.io/nazhang/index.html

TECHNICAL SKILLS

Machine Learning Classi cation, Regression, Clustering, Unsupervised Learning, Sparse Coding Computer Vision Face Morphing/De-Morphing, Morphing Detection, Face Expression Analysis Face Recognition/Detection, Autism Diagnosis, Facial Traits Rating Prediction Data Science Data Engineering (ETL), Data Analysis/Wrangling/Mining/Visualization Deep Learning CNN, GANs, AutoEncoder, Transformer

Frameworks/Tools PyTorch, TensorFlow; Jupyter Lab, Jupyter Notebooks Languages Python, Java, Matlab, C#, C++, C

Python Packages OpenCV/NumPy/Scikit-Learn/Scikit-Image/Pandas/Matplotlib/SciPy/DLIB Database SQL; Oracle DB, MySQL, PostgreSQL

EDUCATION

West Virginia University, WV, US 08/2016 - 05/2023 Ph.D. in Computer Science

Area of Emphasis: Computer Vision, Deep Learning, Image/Video Analysis Beihang University, Beijing, China 09/2009 - 01/2012 Master's Degree in Computer Science

Area of Emphasis: Software Development Process and Management Beijing Information Science and Technology University, Beijing, China 09/2005 - 07/2009 Bachelor's Degree in Computer Science, GPA: 3.8

RESEARCH PROJECTS

Transformer-based Face Morphing and De-Morphing 12/2021 - 09/2022 Research Assistant West Virginia University

Developed a transformer-based scheme for face morphing and de-morphing.

Constructed special losses (face-related/image-related) to learn an optimized latent code of a given face.

Extended the study to face de-morphing based on a reference image.

Demonstrated its superiority to CNN-based morphing methods. Fusion-based Few-Shot Face Morphing Attack Fingerprinting 09/2020 - 11/2021 Research Assistant West Virginia University

Extended morphing attack detection from binary to multiclass morphing attack ngerprinting.

Proposed a few-shot learning framework to learn fusion features of di erent sensor pattern noise by factorized bilinear coding strategy.

Collected a high-resolution Doppelg anger dataset (look-alike face pairs without biological connections).

Extensive experiments showed the outstanding performance. Video-based Facial Micro-Expression Analysis for Autism Diagnosis 10/2022 - 04/2023 Research Assistant West Virginia University

Focused on hour-long interview video analysis by capturing micro-expression of the subjects.

Designed a pipeline to classify autism/control groups by preprocessing, spotting and extracting discrim- inative feature of subtle facial movement via optical ow/attention mechanism/local patch of interest.

Obtained 97.32% accuracy showing the e ciency of acquiring micro-expression for autism diagnosis. Face Dynamics Analysis for Autism Diagnosis on Interview Videos 07/2019 - 08/2020 Research Assistant West Virginia University

Constructed a classi cation system to diagnose autism with di erent severity levels in hour-long videos.

Designed a pipeline to leverage various strategies of 3D spatio-temporal face feature extraction, sparse coding, marginal sher analysis, few-shot learning and scene-level fusion.

Achieved 91.72% accuracy that is comparable to the standardized diagnostic scales. Facial Traits Rating Prediction and Analysis on Autism Participants 12/2020 - 06/2021 Research Assistant West Virginia University

Designed a deep regression model for facial traits rating prediction by transfer learning.

Investigated the di erence between autism and control groups on making social trait judgment. via Layer-wise Relevance Propagation (LRP).

Demonstrated di erent facial areas are involved for di erent social traits judgment between autism and control.

Face Recognition (FR) and Face Quality Analysis 08/2016 - 06/2019 Research Assistant West Virginia University

Explored the impact of di erent face qualities on FR performance by investigating their relationship.

Showed that deep models can allow quality changes to some degrees, but not too large, for the test faces.

Proposed one promising direction for deep learning to further improve its capability in building the relations between face images with large quality gaps.

Gathered a dataset with 356.4K face images of Asian celebrities crawled online after cleaning. WORK EXPERIENCE

Research & Development Engineer 08/2012 - 01/2015

Beijing China-Power Information Technology Co., LTD, Beijing, China

Involved in implementing two large-scale systems (State Grid Infrastructure Control System, State Grid Information Resources System) based on Business Process Management (BPM) Platform using Java, JSP, Oracle and TortoiseSVN, which are successfully deployed and used by State Grid Corporation of China.

Software Engineer Intern 03/2010 - 09/2010

Software Engineering Institute, Beihang University, Beijing, China

Implemented Wireless Network Management System to monitor wireless access points and controllers of Beijing subway transportation wireless network using C# on .Net framework with MySQL. AWARDS

Second Class Scholarship, Beihang University 2010

National Aspiration Scholarship, Beijing, China 2007 Excellent Student Award, BISTU 2009

First Class Scholarship, 5 times, BISTU 2005-2009

Third Class Award of National Physics Contest for College Student, Beijing, China 2007

Contact this candidate