BUILDING A PERSONAL WARDROBE MANAGEMENT APPLICATION

Abstract

Fashion personalization has attracted increasing research attention. However, effective personal clothing management and outfit recommendation remain limited in practical applications, motivating the development of more intelligent and user-oriented solutions. This thesis presents a mobile-based personal wardrobe management system designed to support clothing and outfit organization, while automatically recommending suitable outfits based on the user’s existing wardrobe. The system integrates essential manual management functions with an outfit generation model to optimize outfit selection for different usage purposes. The recommendation component is implemented using a machine learning approach that combines Multi-output Regression and Ranking-based Retrieval. An MLP architecture enhanced with an Attention Pooling mechanism is employed to effectively combine clothing features and generate appropriate outfit combinations. The performance of the proposed model is evaluated using Mean Squared Error and Mean Cosine Similarity as benchmark metrics, showing promising results. To ensure usability and scalability, the system adopts a client–server architecture. The server is developed using NodeJS with a MongoDB database to manage clothing and outfit data and provide APIs for model interaction, while the client is an Android application built with React Native. Functional and user testing are conducted to assess system stability and overall effectiveness.