摘 要:针对传统图像分类方法准确率较低的问题,提出一种基于卷积神经网络和迁移学习思想的图像分类改进方法。利用迁移学习的思想改进卷积神经网络的网络结构及网络参数,然后利用TensorFlow框架实现该模型并对MNIST数据集进行分类,最后将改进卷积神经网络模型的分类准确率与传统分类方法进行对比分析。实验结果表明,改进卷积神经网络模型的分类准确率高达 99.37%,分类性能明显优于其他方法。
关键词:卷积神经网络;迁移学习;TensorFlow;图像分类
DOI:10.19850/j.cnki.2096-4706.2023.05.026
中图分类号:TP183 文献标识码:A 文章编号:2096-4706(2023)05-0109-04
Research on Improved Image Classification Method Based on Convolution Neural Network
SHI Tengfei, SHANG Jiaxiu, WU Zonghang
(School of Electrical Engineering, North China University of Water Resources and Electric Power, Zhengzhou 450045, China)
Abstract: Aiming at the low accuracy of traditional image classification methods, an improved image classification method based on convolution neural network and transfer learning ideas is proposed. Use the idea of transfer learning to improve the network structure and network parameters of the convolution neural network, and then use TensorFlow framework to implement the model and classify the MNIST dataset. Finally, compare the classification accuracy of the improved convolution neural network model with the traditional classification methods. The experimental results show that the classification accuracy of the improved convolution neural network model is up to 99.37%, and the classification performance is significantly better than other methods.
Keywords: convolution neural network; transfer learning; TensorFlow; image classification
参考文献:
[1] 胡貌男,邱康,谢本亮 . 基于改进卷积神经网络的图像分类方法 [J]. 通信技术,2018,51(11):2594-2600.
[2] 张珂,冯晓晗,郭玉荣,等 . 图像分类的深度卷积神经网络模型综述 [J]. 中国图像图形学报,2021,26(10):2305-2325.
[3] 周楠 . 欧阳鑫玉 . 卷积神经网络发展 [J]. 辽宁科技大学学报,2021,44(5):349-356.
[4] 严春满,王铖 . 卷积神经网络模型发展及应用 [J]. 计算机科学与探索,2021,15(1):27-46.
[5] SITAULA C,HOSSAIN M B. Attention-based VGG-16model for COVID-19 chest X-ray image classification [J].Applied Intelligence,2020,51(5):1-14.
[6] 许景辉,邵明烨,王一琛,等 . 基于迁移学习的卷积神经网络玉米病害图像识别 [J]. 农业机械学报,2020,51(2):230-236+253.
[7] 费宁,张浩然 .TensorFlow 架构与实现机制的研究 [J].计算机技术与发展,2019,29(9):31-34.
作者简介:史腾飞(1998—),男,汉族,河南驻马店人,硕士研究生在读,研究方向:控制科学与工程、数据安全。