Using Image Analysis and Unsupervised Learning to identify style patterns in fashion luxury brands

This video presents a concise visual summary of my Master’s thesis, where I explore how Artificial Intelligence can transform the way luxury e-commerce organizes and displays products. The research investigates whether Computer Vision and Unsupervised Learning—specifically Vision Transformer (ViT) embeddings combined with Agglomerative Clustering—can visually segment luxury eyewear into coherent aesthetic groups that integrate both style and price. Using a dataset of 763 images scraped from six major luxury brands (Dolce & Gabbana, YSL, Prada, Fendi, Bottega Veneta, and Cartier), the project builds a deep-learning pipeline to analyze how shape, design language, and pricing correlate across the industry. The results show that ViT embeddings capture high-level aesthetic patterns beyond color or material, revealing visual identities consistent with each brand’s positioning. By merging visual and economic dimensions, the framework helps reduce consumer choice overload and provides brands with new tools for competitive benchmarking and assortment optimization. This video highlights the key concepts, methods, and insights behind the thesis—summarizing how AI can bridge data intelligence and aesthetic perception in the digital era.