Multi-Label Image Recognition Algorithm Based on Convolutional Block Attention Module

This paper proposes a multi label image recognition algorithm based on CBAM (Convolutional Block Attention Module) to address the issue of partial feature neglect in multi label images. The algorithm constructs a co-occurrence matrix between labels using a graph convolutional neural network, and uses a ResNet-34 network based on CBAM as an improved convolutional neural network to form a new recognition network.