OutfitX: A Deep Learning Framework for Personalized Outfit Recommendations

The expansion of e-commerce and custom fashion has made the personalization of outfit recommendation systems crucial. Outdated approaches use manual efforts or basic rules, which do not suit specific users and their contexts. The author has worked on and proposed Outfit-X, a personalized outfit recommendation system powered by deep learning (ResNet50 classification). Using convolutional neural networks (CNNs) and graph neural networks (GNNs) along with metadata fusion enables the system to give contextually relevant outfit suggestions. The user takes a picture of each clothing item with a digital wardrobe application, and the application suggests suitable items (e.g., shoes, bags) in images that match them visually and semantically. Evaluations conducted using the DeepFashion dataset revealed 94.2 % accuracy in compatibility prediction, exceeding baseline models such as Siamese Networks and collaborative filtering. The system integrates explainable artificial intelligence (XAI) to clarify the matching attributes in a recommendation, for example, matching colors and harmonious patterns, which helps users trust the system more.