A Study on Fashion and Style Recommendation System  using Machine Learning and Deep Learning Algorithms

Abstract : A fashion recommendation system is a computer-driven solution that suggests clothing and accessories based on visual features extracted from images. People can choose cloths or other fashion accessories from a wide variety of products. The rise of e-commerce companies has significantly changed people’s shopping style. Shoppers can pick out clothes and accessories from a wide variety of products, all matched to what they like, without getting off the couch. In this ever-changing world of online shopping, systems that suggest fashion and style ideas are key to making the experience better for users. These systems offer personalized choices based on what each person likes, what's trendy, and what fits the occasion. They don't just save time - They help users find unique styles that match their personality, making online shopping simple and enjoyable. This paper presents an AI-based fashion recommendation system that enables users to upload images of fashion items and receive similar product recommendations and style suggestions. The system utilizes deep learning-based feature extraction with MobileNet, a pre-trained convolutional neural network (CNN), to generate embeddings that capture visual attributes such as shape, color, and texture. These embeddings are then processed using k-Nearest Neighbours (k-NN) to find the most visually similar fashion items from a dataset. Additionally, a complementary style recommendation module suggests outfit pairings based on predefined category relationships. Unlike traditional recommendation systems that rely on user browsing history or purchase patterns, this approach is purely image-driven and requires no prior user data. The system is designed for fashion e-commerce platforms, mobile applications, and virtual styling assistants, offering an efficient, scalable, and user-friendly solution for personalized fashion discovery.

Keyword MobileNet, Machine learning techniques, Nearest neighbour (k-NN), Deep learning, and Clothing Categories.