摘要
文章讲述卷积神经网络的思想、基础知识,介绍了经典的卷积神经网络的方法,引入了卷积神经网络的最新改进方法,并对各个模型架构进行了分析,最后,通过实验,展示了多种方法性能,为研究者选择模型进行科研和教学提供了依据。
This paper traces the idea and basic knowledge of convolutional neural network,introduces the classical methods of convolutional neural network,as well as the latest improved methods of convolutional neural network,and analyzes the architecture of each model.Finally,through experiments,the performance of various methods is demonstrated,which provides a basis for researchers to choose models for scientific research and teaching.
作者
郭俊亮
张洪川
GUO Junliang;ZHANG Hongchuan
出处
《科技创新与应用》
2021年第23期16-18,22,共4页
Technology Innovation and Application
基金
铜仁市科学技术局基础科学研究项目(铜市科研(2020)118号)。
关键词
图像识别
卷积神经网络
经典模型
性能改进
image recognition
convolution neural network
classical model
performance improvement