Convolutional Neural Network-Based Garment Classification Using Fashion MNIST Dataset—A Comparative Analysis

The surge in online wear shopping has made it imperative for e-commerce platforms such as Myntra, Ajio, Amazon, Flipkart, and Meesho to accurately categorize the wide array of wearable items available on their websites. Addressing this challenge requires efficient image classification. Convolutional Neural Networks (CNNs), a specialized class of deep neural networks, excel in extracting specific features from images, making them ideally suited for image classification tasks. Accurate classification of garments is essential for enhancing search results and improving the overall shopping experience for customers. Manual categorization and description of clothing items can be arduous and time-consuming, particularly for retailers managing extensive product catalogs. The integration of CNNs not only enhances the customer shopping experience but also brings cost and time savings for both retailers and customers. In our study, we conducted an analysis of four prominent CNN models—Lenet, AlexNet, VGGNet, and ResNet—using the Fashion MNIST Dataset. Additionally, we performed a comparative study among these models, evaluating their respective strengths, weaknesses, architectures, accuracy rates, and determining which among them is most suitable for classification purposes.