DESIGN OF A CLOTHING E-COMMERCE WEBSITE WITH IMAGE-BASED SEARCH

Abstract

In the current 5.0 era, the explosion of e-commerce has significantly changed customer shopping habits. However, finding the products they want among hundreds or thousands of items remains a significant challenge. Traditional keyword-based search often has limitations with fashion products, where visual characteristics such as style and color are difficult to describe in words. To solve this problem, the thesis proposes building an e-commerce fashion website system that integrates image-based search functionality. The highlight of this thesis is the application of the CLIP (Contrastive Language-Image Pre-training) model, fine-tuned on the DeepFashion dataset using Proxy Anchor Loss, which helps the model increase its ability to extract features and match similarities. Besides the image-based search feature, the system also provides a complete ecommerce platform with all the essential functions to ensure a smooth user experience from ordering and payment to automated support via Chatbot. It also includes full tools for business management, inventory, and marketing. The result is a modern, user-friendly fashion e-commerce platform. It connects the visual shopping experience with traditional online retail, giving customers a way to discover products through images while maintaining full e-commerce functionality.