Clothing classification algorithm based on convolution and Transformer fusion

Abstract

Aiming at the problem that the traditional clothing classification algorithm based on convolutional neural network cannot meet the needs of massive and diverse clothing classification, a clothing classification network with convolutional attention fusion is proposed. The network adopts a parallel structure, including a ResNet branch and a Transformer branch, which makes full use of the local features extracted by convolution operation and the global features extracted by self-attention mechanism to enhance the representation learning ability of the network, thereby improving the performance and generalization ability of clothing classification algorithm. To verify the effectiveness of this method, comparative experiments were carried out on the datasets Fashion-MNIST and DeepFashion. The results show that on the dataset Fashion-MNIST, the method achieved an accuracy of; on the dataset DeepFashion, the method achieved an accuracy of; the experimental results of this method are better than those of other comparative methods.