High-precision Image Classification Algorithm based on Attention Mechanism and Multi-scale Features

Abstract

High-precision image classification has steadily emerged as a key area of research interest due to the extensive use of image classification technologies in many different domains. The study enhances the conventional feature pyramid networks (FPN) and suggests a high-precision image classification model in an attempt to further increase the precision and effectiveness of picture classification. The model enhances the ability of convolutional neural network (CNN) to focus on key information by combining the channel attention and spatial attention mechanisms. The outcomes indicated that the improved CNN model achieved 77.50% classification accuracy on the ImageNet dataset and 94.20% on the CIFAR-10 dataset, which was significantly higher than the control model. In addition, in the classification of different types of high-precision images, the improved CNN model performed well in the recall, F1 score, and robustness metrics. Their values were 94.3%, 94.6%, and 93.5%, respectively. The results show that the high-precision image classification model is able to capture the key features and detail information in the image more effectively, which significantly improves the classification accuracy and robustness. This study provides a new technical tool for high-precision image classification tasks.