Date of Graduation

5-2025

Document Type

Thesis

Degree Name

Bachelor of Science in Data Science

Degree Level

Undergraduate

Department

Data Science

Advisor/Mentor

Schubert, Karl

Committee Member

Tran, Quoc

Second Committee Member

Plummer, Sean

Abstract

With the rapid expansion of e-commerce over time, ensuring the diversity and quality of product images has become a critical challenge infeasible for human completion. In conjunction with Walmart Global Tech for the Team 1 Data Science Practicum Project, image classification models were trained to assess product image sets, but training and deploying such models often involves repetitive code and inefficient processes. This thesis presents a reusable modeling library, named IcyLib, designed to streamline the training, validation, and testing of image classification models as well as dataset importation using PyTorch. IcyLib provides a structured yet flexible approach for model implementation, currently supporting ResNet and EfficientNet architectures with minimal user effort through JSON configuration files. By automating dataset handling, model training, and performance evaluation, IcyLib significantly reduces redundancy and potential errors while improving usability, reproducibility, and readability of code. Experimental results demonstrate the efficiency and efficacy of this framework in classifying Walmart.com product images, offering a scalable solution for future applications in e-commerce image analysis.

Keywords

e-commerce; image classification; automated machine learning

Download

Included in

Data Science Commons

Share

COinS