Image classification is one of the core tasks in the field of computer vision. The rapid development of deep learning provides new possibilities for improving the performance of image classification algorithm. However

the existing deep learning models still face the problems of high calculating cost and insufficient generalization ability when dealing with complex image classification tasks. This paper takes the deep learning image classification algorithm as the research object and combines mainstream models such as convolutional neural network (CNN) and Transformer to propose an optimization method based on data enhancement

model compression and attention mechanism. Through theoretical analysis and technical improvement

this paper provides a new idea for the optimization of image classification algorithm and lays a foundation for its practical application.