基于优化改进的Xception模型的垃圾图片分类  

Garbage Image Classification Based on An Optimized and Improved Xception Model

作　　者：刘后胜 张洋 陶健林[1] Liu Housheng;Zhang Yang;Tao Jianlin(School of Information Technology,Anqing Vocational and Technical College,Anqing 246003,China)

机构地区：[1]安庆职业技术学院信息技术学院,安徽安庆246003

出　　处：《黄山学院学报》2022年第3期30-32,共3页Journal of Huangshan University

基　　金：安徽省教育厅质量工程项目(2020kfkc310,2020jyxm1127)。

摘　　要：Xception模型的训练精度很高,但预测精度不是很高,即出现模型训练过拟合现象,文章对该模型结构做了进一步的优化改进,将改进后的模型应用于垃圾图片分类,实验结果显示优化改进的模型虽然训练精度有所降低,但预测精度比原始的Xception模型有所提高,在一定程度上纠正了过拟合现象。The training accuracy of Xception original model is very high, but the prediction accuracy is relatively low, showing model training overfitting. This paper further optimizes and improves the model structure, and applies the improved model to garbage image classification. The experimental results show that although the training accuracy of the optimized and improved model is reduced, the prediction accuracy is improved compared with the original Xception model, and the overfitting phenomenon is corrected to a certain extent.

关 键 词：Xception模型 优化 过拟合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]