摘要
基于Keras深度学习框架和卷积层取反操作,提出一种改进型的卷积神经网络结构,网络结构首层采用卷积层取反以增加有效特征信息的传递,有效结合Leaky ReLU激活函数传递至下一层,最后采用Softmax分类器实现图像分类。在两个公共数据集上,同传统的卷积神经网络模型做对比实验,实验结果表明,改进的卷积网络模型是有效的。
Based on Keras deep learning framework and convolution layer inverse operation,an improved Convolutional Neural Network structure is proposed in this paper.The first layer of the network structure uses convolutional layer inversion to increase the transmission of effective feature information.The LeakyReLU activation function is effectively combined to the next layer.Finally,the Softmax classifier is used to implement image classification.Compared with the traditional Convolution Neural Network model on two common datasets,the experimental results show that the improved convolution network model in this paper is effective.
作者
张斌
王强
ZHANG Bin;WANG Qiang(College of Computer Science,Chengdu University of Information Technology,Chengdu 610225,China;State Grid Sichuan Electric Power Company Deyang Power Supply Company,Deyang 618000,China)
出处
《成都信息工程大学学报》
2019年第1期39-43,共5页
Journal of Chengdu University of Information Technology
关键词
深度学习
卷积神经网络
图像分类
激活函数
deep learning
Convolutional Neural Network
image classification
activation function