{"title":"基于卷积神经网络的葡萄叶片病害分类","authors":"Khaing Zin Thet, Khine Khine Htwe, M. Thein","doi":"10.1109/ICAIT51105.2020.9261801","DOIUrl":null,"url":null,"abstract":"Most of the grape plant diseases are severe and spread rapidly. The grape plant diseases start on the leaf and then spread to stem, fruit, and root. It has problems with time-consuming and lacking knowledge for farmers in remote areas to classify grapes leaf diseases because of limited access to human experts in developing countries. In recent research, grape leaf diseases are classified with fine-tuning VGG16 Network. But the classification was not received good accuracy results with VGG16 Architecture. This system proposes transfer learning via fine-tuning of VGG16 network, one of CNN Architecture, to classify diseases on grape leaf The system used Global Average Pooling (GAP) layer instead of VGG16's two fully connected layers before final classification SoftMax layer to improve accuracy result of fine-tuning VGG16 for grape leaf diseases classification. The proposed system mainly analyzed healthy leaves and five leaves diseases, named anthracnose, downy mildew, black measles, isariopsis leaf spot, nutrient insufficient, on 6000 images dataset of Myanmar Grapevine Yard. The proposed system compared the accuracy with VGG16 using fully connected layers and VGG16 using SVM classifier. The proposed system outperformed with 98.4% accuracy than others.","PeriodicalId":173291,"journal":{"name":"2020 International Conference on Advanced Information Technologies (ICAIT)","volume":"101 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"15","resultStr":"{\"title\":\"Grape Leaf Diseases Classification using Convolutional Neural Network\",\"authors\":\"Khaing Zin Thet, Khine Khine Htwe, M. Thein\",\"doi\":\"10.1109/ICAIT51105.2020.9261801\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Most of the grape plant diseases are severe and spread rapidly. The grape plant diseases start on the leaf and then spread to stem, fruit, and root. It has problems with time-consuming and lacking knowledge for farmers in remote areas to classify grapes leaf diseases because of limited access to human experts in developing countries. In recent research, grape leaf diseases are classified with fine-tuning VGG16 Network. But the classification was not received good accuracy results with VGG16 Architecture. This system proposes transfer learning via fine-tuning of VGG16 network, one of CNN Architecture, to classify diseases on grape leaf The system used Global Average Pooling (GAP) layer instead of VGG16's two fully connected layers before final classification SoftMax layer to improve accuracy result of fine-tuning VGG16 for grape leaf diseases classification. The proposed system mainly analyzed healthy leaves and five leaves diseases, named anthracnose, downy mildew, black measles, isariopsis leaf spot, nutrient insufficient, on 6000 images dataset of Myanmar Grapevine Yard. The proposed system compared the accuracy with VGG16 using fully connected layers and VGG16 using SVM classifier. The proposed system outperformed with 98.4% accuracy than others.\",\"PeriodicalId\":173291,\"journal\":{\"name\":\"2020 International Conference on Advanced Information Technologies (ICAIT)\",\"volume\":\"101 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-11-04\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"15\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 International Conference on Advanced Information Technologies (ICAIT)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICAIT51105.2020.9261801\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 International Conference on Advanced Information Technologies (ICAIT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICAIT51105.2020.9261801","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 15
摘要
葡萄植株病害大多严重,传播迅速。葡萄植株的病害从叶片开始,然后蔓延到茎、果实和根。由于发展中国家与人类专家的接触有限,偏远地区的农民在分类葡萄叶病方面存在耗时和缺乏知识的问题。在最近的研究中,葡萄叶片病害采用微调VGG16网络进行分类。但使用VGG16架构进行分类并没有得到很好的准确率结果。本系统提出通过对CNN架构之一的VGG16网络进行微调迁移学习,对葡萄叶片病害进行分类,在最终分类SoftMax层之前,系统使用Global Average Pooling (GAP)层代替VGG16的两个完全连通层,以提高微调VGG16对葡萄叶片病害分类的准确率结果。该系统主要对缅甸葡萄场6000幅图像数据集上的健康叶片和5种叶片病害进行分析,分别为炭疽病、霜霉病、黑麻疹、沙蒿叶斑病、营养不足病。该系统与使用全连接层的VGG16和使用SVM分类器的VGG16的准确率进行了比较。该系统以98.4%的准确率优于其他系统。
Grape Leaf Diseases Classification using Convolutional Neural Network
Most of the grape plant diseases are severe and spread rapidly. The grape plant diseases start on the leaf and then spread to stem, fruit, and root. It has problems with time-consuming and lacking knowledge for farmers in remote areas to classify grapes leaf diseases because of limited access to human experts in developing countries. In recent research, grape leaf diseases are classified with fine-tuning VGG16 Network. But the classification was not received good accuracy results with VGG16 Architecture. This system proposes transfer learning via fine-tuning of VGG16 network, one of CNN Architecture, to classify diseases on grape leaf The system used Global Average Pooling (GAP) layer instead of VGG16's two fully connected layers before final classification SoftMax layer to improve accuracy result of fine-tuning VGG16 for grape leaf diseases classification. The proposed system mainly analyzed healthy leaves and five leaves diseases, named anthracnose, downy mildew, black measles, isariopsis leaf spot, nutrient insufficient, on 6000 images dataset of Myanmar Grapevine Yard. The proposed system compared the accuracy with VGG16 using fully connected layers and VGG16 using SVM classifier. The proposed system outperformed with 98.4% accuracy than others.