摘要
为了提高蜡染纹样的分类准确率,提出一种改进的VGGNet分类模型,将最后池化层的输出进行全局平均池化后直接与分类神经元进行全连接。采用数据增强技术扩充训练集样本数量,提高模型的泛化能力,使用迁移学习方法将预先训练好的VGGNet参数作为初始化参数,提高模型训练的效率。实验结果表明,该方法能够有效提高蜡染纹样的分类准确率。
To improve the classification accuracy of indigo batik pattern,a classification model based on the improved convolution neural network model of VGGNet is proposed,the outputs of the final pooling layer are pooled by global average and connected directly with the classification neurons. Data enhancement technology is used to expand the number of training samples and improve the generalization ability of the model,using transfer learning method,the pre trained VGGNet parameters are used as initialization parameters to improve the efficiency of model training. Experimental results show that this method can effectively improve the classification accuracy of indigo batik pattern.
作者
周长敏
佘佐明
吴安丽
ZHOU Changmin;SHE Zuoming;WU Anli(School of Big Data Engineering,Kaili University,Kaili 556011;School of Economics and Management,Kaili University,Kaili 556011;School of Art and Design,K aili University,Kaili 556011)
出处
《现代计算机》
2021年第22期117-121,共5页
Modern Computer
基金
贵州省教育厅2019年青年科技人才成长项目(黔教合KY字[2019]194)。
关键词
蜡染纹样分类
卷积神经网络
迁移学习
VGGNet
Classification of Indigo Batik Pattern
Convolutional Neural Networks
Transfer Learning
VGGNet
