Multi-Attribute Clothing Classification Using Modular Deep Learning Models: A ResNet-Centric Architecture

Abstract

Translate Article

Automatic clothing labelling in the e-commerce industry is an emerging field where deep learning techniques are proving invaluable. This research explores the use of Convolutional Neural Networks (CNNs) for classifying and categorizing clothing items. The study employs a modular ResNet-based approach, trained on the DeepFashion and Fashion Product Images datasets, to create specialized models for various attributes such as category, colour, gender, and additional clothing features. The models demonstrate strong performance, with accuracies ranging from 82% to 95%, indicating their potential for improving automated product recognition in e-commerce.