PSBCNN : Fine-grained image classification based on pyramid convolution networks and SimAM

Fine-grained image classification has been an important but challenging task due to high intra-class variances and low inter-class variances. Therefore, we propose a fine-grained image classification algorithm based on pyramidal convolutional neural network and 3-D attention module to solve these problems. The algorithm captures different levels of detail in the scene and assigns greater weights to the distinguishing features, allowing the neural network model to focus more on local regions and thus improve the accuracy of fine-grained classification. Qualitative experiments on three benchmark fine-grained datasets demonstrate the superiority of our proposed method.