Machine Learning, computer vision, deep learning, data visualization

Yushuo Niu

Æ 860-***-**** [ ******.***@*****.*** linkedin.com/in/yushuo-niu-088522a4

Graduate research assistant specializing in computer vision and machine learning, with extensive experience in change detection, object detection, and data-driven dynamical systems. Looking for a full-time machine learning engineer and computer vision engineer position. Education

University of Connecticut (UCONN) Storrs, CT

Ph.D. Computer Science and Engineering; GPA: 3.75/4.00 Expected August 2025 University of Connecticut (UCONN) Storrs, CT

M.S. Mechanical Engineering October 2016

Shandong University (SDU) Shandong, China

B.S. Mechanical Design, Manufacturing, and Automation July 2012 Technical Skills

Programming: Python, C/C++, Java, SQL, MATLAB, R, Vim, Latex, CUDA, Git, Bash Software: Pytorch, Tensorflow/Keras, Scikit-learn, OpenCV, Matplotlib, PySINDy, Pandas, Docker, Linux Professional Area: Deep Learning, Computer Vision, Few-shot Learning, Statistical Learning, data-driven dynamical systems

Certification: Deep Learning Specialization, Generative Adversarial Networks (GANs) Specialization

(Deeplearning.ai), IBM Data Science Professional Certificate Research Experience

Semi-Siamese Neural Networks for Robust Defect Detection Storrs, CT Research Assistant Jun 2019 – Present

• Proposed an innovative Semi-Siamese architecture leveraging few-shot learning to achieve real-time high-resolution defect detection in binder jet 3D printing, uniquely robust against variations in camera settings and lighting conditions.

• Combined with data augmentation, dramatically reducing the requirement for labeled data to just 57 pairs of experimental images.

• Achieved a 54% reduction in model training time compared to available methods, attributed to the lightweight architecture, significantly enhancing training efficiency.

• Outperformed state-of-the-art methods by 8% F1 score on the benchmark Synthetic-Aperture Radar (SAR) dataset.

Physics-Informed Image Segmentation of Filamentous Carbon from in-situ TEM Storrs, CT Research Assistant Apr 2023 – present

• Developed a new Siamese neural network model to predict changes in the filament carbon over time from Transmission Electron Microscopy (TEM) video frames.

• Proposed an innovative loss function that integrates physics-informed regularization, enhancing the prediction accuracy of small features by 11%.

Data-Driven Learning of Dislocation Evolution in Shocked Metallic Microstructure Storrs, CT Research Assistant Mar 2023 – present

• Applied the sparse dictionary-based SINDy algorithm for data-driven learning of evolutionary equations describing a system of three types of dislocations—Perfect, Shockley, and Locks (immobile dislocations)—in a single crystal aluminum system subjected to varying impact velocities.

• Developed a novel approach for learning an adaptive family of evolutionary equations, enabling extrapolation across impact velocities and extracting physical insights into dislocation evolution. Pneumonia Detection using Mask-RCNN and YOLOv3 on Chest X-Rays Storrs, CT Research Assistant Aug 2019 – Dec 2019

• Reprocessed the images for optimal performance in deep learning models, which included normalization, resizing, and augmentation.

• Implemented and customized Mask R-CNN and YOLOv3 models for the specific task of detecting pneumonia-related features in chest X-rays (CXRs).

• Analyzed the model’s predictions to ensure clinical relevancy and interoperability, highlighting areas of CXRs indicative of pneumonia, and achieved an F1 score of 0.73. Acoustic Signal Spread-spectrum System Using Piezoelectric Transducer Storrs, CT Graduate Student Jan 2015 – Sep 2016

• Developed MATLAB code to apply the KLM (Krimholtz et al.) model to characterize Piezoelectric Transducer’s electrical and mechanical behavior.

• Implemented Manchester coding to translate the baseband signal to a high-frequency signal to realize acoustic signal Spread-spectrum with piezoelectric transducers. Awards & Achievements

Computer Science & Engineering First Place Award UConn 10th Annual Graduate Poster Competition. (Feb 2024)

First Prize of Camera Group 6th National Undergraduates Freescale Cup Intelligent Car Racing (Aug 2010) Third Prize 3rd National University Students Social Practice and Science Contest on Energy Saving & Emission Reduction “CELARTEM EN-FORTECH”. (Aug 2010)

Presentations

Niu, Y., Chadwick, E., Ma, A.W. and Yang, Q., 2024, June. Robust Defect Detection for Binder Jet 3D Printing with Semi-Siamese Neural Networks. Contributed talk, AI Innovations, TechConnect World Innovation 2024, Washington, DC. Niu, Y., Chadwick, E., Ma, A.W. and Yang, Q., 2024 February. Robust Defect Detection for Binder Jet 3D Printing with Semi-Siamese Neural Networks. Invited Speaker, Nondestructive Inspection for AM parts (NDIxAM) 2024, New Jersey, NJ.

Niu, Y., Chadwick, E., Ma, A.W. and Yang, Q., 2021, October. Machine Learning In-Process Defect Detection and Correction. SHAP3D Center, Atlanta, GA

Niu, Y., Chadwick, E., Ma, A.W. and Yang, Q., 2020, May. Machine Learning-Enabled In-Process Defect Detection. SHAP3D Center, Lowell, MA

Posters

Niu, Y. Extrapolation in Data-Driven Learning of Complex Reactive Processes. Society for Industrial and Applied Mathematics (SIAM) Conference on Mathematics of Data Science (MDS24). Moaddel, T., Bertola, P.S.R., Niu, Y., Quan C., Vinski, P., Dasgupta, B., Ghatlia, N., Shiloach, A., Yang, Q., Ma, A.W., 2024, March, Sub-micron emulsion based mild bodywash formulations for superior active deposition. American Academy of Dermatology Association (AAD) 2024.

Chadwick, E., Chang, S.Y., Pardakhti, M., Tan, M., Niu, Y., Yang, Q., Kang, S.Y., Chaudhuri, B., Ma, A.W., 2021, October. Inkjet-based 3D Printing of Functional Materials. Advanced Manufacturing Innovation, TechConnect World 2021, Washington, DC.

Publications

Niu, Y., Chadwick, E., Ma, A.W. and Yang, Q., 2023, September. Semi-Siamese Network for Robust Change Detection Across Different Domains with Applications to 3D Printing. In International Conference on Computer Vision Systems

(pp. 183-196). Cham: Springer Nature Switzerland.

Master’s Thesis, University of Connecticut, Niu, Y., 2015. Acoustic Signal Spread-spectrum System Using Piezoelectric Transducer.

Niu, Y., Li, T., Bin, M., Zhu, Y., Yang, Q., Deep Few-Shot Learning for Filamentous Carbon Segmentation from in-situ TEM. submitted.

Niu, Y., Sebastian, R., Dongare, A., Yang, Q., Data-driven Modeling of Dislocation Evolution in a Shocked Metallic Microstructure. in preparation.

Contact this candidate