Fashion Recommendation System using ResNet50

This paper presents a new approach towards fashion recommendation systems, exploiting deep learning techniques—especially the Residual Network (ResNet) architecture—to mine complex and essential information out of images. In this way, by employing Convolutional Neural Networks to analyze fashion images, the distance between user preferences and the available product catalog will narrow significantly, providing a much more personalized and relevant shopping experience. This line of work considers classical recommendation system shortcomings, which mostly end up not being able to reflect the deeply subjective and very complicated nature of fashion decision-making. The system advocates the application of ResNet50 for the extraction of features and uses the nearest neighbors algorithm in making recommendations with the aim of increasing relevance and accuracy of the suggestions while giving a completely good user shopping experience. Through these comparisons, the results of the experiment proved the effectiveness of the system in authentically giving the customer an individualized, enjoyable shopping experience and the effectiveness in applying deep learning to develop the fashion recommendation system.