关键词: 深度学习, 图像分类, 高低频学习, 傅里叶变换

Abstract: At present,deep learning has achieved considerable success in many fields. In recent years,as a representative task in the field of deep learning,image classification tasks have made great progress. In traditional deep learning classification algorithms,data-driven end-to-end training is generally adopted,that is,images in RGB space are directly input into the deep neural network,and the deep neural network is continuously optimized through the gradient descent algorithm, satisfactory classification accuracy has been achieved gradually. Images can be decomposed into high-frequency information and low-frequency information in general cognition. However,in traditional deep neural network algorithms,the separate role of the high-frequency and low-frequency information of the image in the training process is not considered. So,Fourier transform to extract the high and low frequency information of the image,combined with the course learning method,proposes a deep classification method based on progressive high and low frequency. This method is a brand-new deep learning training method,which can improve the performance of the deep neural network without changing the size and parameter amount of the deep neural network model.The current image classification task is compared with the current mainstream methods,and a large number of experiments are carried out on the benchmark data set. The experimental results confirm the effectiveness of this method.

Key words: deep learning, image classification, high and low frequency learning, Fourier transform