Convolutional Network Model using Hierarchical Prediction and its Application in Clothing Image Classification

This article proposes a kind of new hierarchical classification model (HCNN) based on convolutional networks. The coarse category-fine category classification layer is set from the lower layer to the upper layer of the network. The experiment proves the necessity of hierarchical classification structure to classify clothing images. In previous studies, when using convolutional neural networks for image classification or other machine learning methods, most of them did not consider hierarchical. This paper attempts to classify apparel datasets using hierarchical CNN for the first time. The suggested model is a knowledge-embedded classifier which conveys some hierarchical information. We implemented HCNN using VGGNet as the underlying framework on the Fashion-MNIST dataset. The results show that the loss is reduced and the accuracy is improved when compared with the base model without hierarchy.