基于改进VGG-16神经网络的图像分类方法
投稿时间:2020-10-12  修订日期:2020-11-09  点此下载全文
引用本文:
摘要点击次数: 32
全文下载次数: 0
作者单位E-mail
田佳鹭 沈阳师范大学 17767711368@163.com 
邓立国 沈阳师范大学  
基金项目:教育信息化云生态环境的架构大数据研究(JG16DB395)
中文摘要:为提高图像分类模型的准确度,提出一种迁移学习VGG-16并对其进行改进的图像分类方法,即NewVGG-16模型。首先从ImageNet数据集中选取十种不同类型的部分图像数据,进行去噪、标准化等预处理;接着迁移学习VGG-16模型同时将其改进,模型的优化包括改进池化层为sort_pool2d,在每个卷积层后面添加BN层以增强规范性,并选用Adaboost分类器提升整体的分类性能。通过训练集实现模型参数的调整,用测试集检验其准确性。实验证明,该模型能有效提升图像分类的准确性和适用性,准确度可达到98.75%。
中文关键词:VGG-16  卷积神经网络  图像分类  迁移学习  Adaboost
 
Image Classification Method Based on Improved VGG-16 Neural Network
Abstract:In order to improve the accuracy of image classification model, a new image classification method based on transfer learning vgg-16 is proposed. Firstly, ten different types of partial image data are selected from Imagenet data set for preprocessing such as denoising and standardization; then, vgg-16 model is transferred and improved at the same time. The optimization of the model includes improving the pooling layer to sort_ The BN layer is added after each convolution layer to enhance the normalization, and AdaBoost classifier is used to improve the overall classification performance. The model parameters are adjusted by training set, and the accuracy is tested by test set. Experiments show that the model can effectively improve the accuracy and applicability of image classification, and the accuracy can reach 98.75%.
keywords:VGG-16  convolutional neural network  image classification  transfer learning  Adaboost
查看全文   查看/发表评论   下载pdf阅读器