Comparison of Different Models for Clothing Images Classification Studies

Abstract

In this paper, we employed three machine learning models, i.e., pure 5-layer convolutional neural network, VGG-16 model, and XGBoost algorithm model. We trained and tested these models on the Fashion-MNIST dataset. By comparing the classification accuracy and the training time, the results show that 1) the pure convolutional neural network is a very effective method for clothing images classification; 2) The complex structure of VGG 16 increases the training time, and the risk of overfitting; 3) XGBoost does not show the efficiency benefits of multi-threading on this issue.