Image classification algorithm based on deep learning
Jing Zhao, Lichun Wan · 2024
The focus of this paper is to solve the problem of increasing training errors and decreasing image classification performance of Convolutional Neural Network (CNN) with increasing network depth. CNN is a deep feedforward neural network model with convolutional structure, high fault tolerance and efficient computing ability, and has been widely used in image classification and other fields. Although the fitting ability of CNN is enhanced with increasing network depth, higher training errors are produced with increasing depth. Therefore, we will explore how to solve this problem. This study is based on the Residual Networks (ResNet) model, focusing on its performance in image classification at different depths. Experiments show that on the CIFAR-10 and CIFAR-100 datasets, the classification error rate generally decreases with the increase of ResNet model depth. Among them, the performance is best on the CIFAR-10 dataset. Additionally, this study applies the Random Erasing (RE) data augmentation algorithm on this basis. This algorithm offers significant advantages and serves the purpose of deep learning. Through model analysis and leveraging advanced technical methods, it greatly enhances the precision of image information acquisition, enabling better classification of different types and increasing the model's robustness to occlusion.