DEEP LEARNING-BASED VISUAL STYLE MATCHING: A CNN-DRIVEN APPROACH FOR FASHION RECOMMENDATIONS

The growing demand for personalized fashion recommendations has driven advancements in artificial intelligenceand computer vision. This study proposes a deep learning-based framework for style matching, leveraging theVGG16 Convolutional Neural Network (CNN) to extract rich visual features from fashion images. Unlikeconventional recommendation systems that rely on textual descriptions or metadata, our approach focuses purelyon image-based analysis, ensuring accurate and visually coherent outfit suggestions. The system computes featuresimilarity using cosine distance, enabling precise retrieval of fashion items that align with individual stylepreferences. Extensive experimentation demonstrates the model’s ability to recognize aesthetic compatibility withhigh accuracy, outperforming traditional heuristic-based methods. By integrating deep learning with contentbased image retrieval, this framework enhances digital wardrobe curation, fashion e-commerce, and personalizedstyling applications. Future work aims to incorporate multimodal learning, real-time trend adaptation, andimproved contextual awareness for an even more refined recommendation experience.