Image Classification and Visual Search for Fashion Items Using Deep Learning Algorithms

This research was conducted to improve the efficiency of ecommerce by improving the classification and visual search of fashion products’ images utilizing learnings from deep learning. The research utilized the VGG19 model for classification which achieved high accuracy, precision, recall, and an impressive AUC score. The results, therefore, demonstrate that the VGG19 model is capable of capturing the small features that are highly useful in classification. Furthermore, a CNN autoencoder integrated with ResNet-50 was used for visual search which retrieved visually similar images proving practical approach to enhancing the use’s convenience of e-commerce platform.