{"title":"基于VGG16模型的景观图像分类迁移学习","authors":"Abubakar MAYANJA, İlker Ali ÖZKAN, Şakir TAŞDEMİR","doi":"10.58190/icat.2023.20","DOIUrl":null,"url":null,"abstract":"In recent times, the need for the use of image classification techniques of machine learning to solve worldly problems in various areas such as agriculture, the health sector, and tourism is rocketing up day by day. Traditionally, one of the most used techniques in image classification is the use of deep neural networks called convolution neural networks (CNN). To come up with a good network model, one needs to have an enormous quantity of data in the form of images and design a network model from scratch in a trial-and-error way. This not only takes a lot of time but also requires very powerful computation equipment such as graphical processing units (GPU). To overcome such barriers, a machine learning technique called transfer learning enables the use of already trained network models in the form of fine-tuning them to solve related issues. In this work, the 2014 ImageNet winner model called Vgg16 was adopted to classify landscape images in the Intel dataset. The dataset contains 5 categories of images namely buildings, forest, glacier, mountain, sea, and street. The performance of Vgg16 was compared to that of a 7-layer ordinary convolution neural network and the results showed that transfer learning with Vgg16 outperformed the ordinary network by 90.1% for Vgg16 compared to 62.5% for the ordinary convolutional neural network model.","PeriodicalId":20592,"journal":{"name":"PROCEEDINGS OF THE III INTERNATIONAL CONFERENCE ON ADVANCED TECHNOLOGIES IN MATERIALS SCIENCE, MECHANICAL AND AUTOMATION ENGINEERING: MIP: Engineering-III – 2021","volume":"60 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Utilizing Transfer Learning on Landscape Image Classification Using the VGG16 Model\",\"authors\":\"Abubakar MAYANJA, İlker Ali ÖZKAN, Şakir TAŞDEMİR\",\"doi\":\"10.58190/icat.2023.20\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In recent times, the need for the use of image classification techniques of machine learning to solve worldly problems in various areas such as agriculture, the health sector, and tourism is rocketing up day by day. Traditionally, one of the most used techniques in image classification is the use of deep neural networks called convolution neural networks (CNN). To come up with a good network model, one needs to have an enormous quantity of data in the form of images and design a network model from scratch in a trial-and-error way. This not only takes a lot of time but also requires very powerful computation equipment such as graphical processing units (GPU). To overcome such barriers, a machine learning technique called transfer learning enables the use of already trained network models in the form of fine-tuning them to solve related issues. In this work, the 2014 ImageNet winner model called Vgg16 was adopted to classify landscape images in the Intel dataset. The dataset contains 5 categories of images namely buildings, forest, glacier, mountain, sea, and street. The performance of Vgg16 was compared to that of a 7-layer ordinary convolution neural network and the results showed that transfer learning with Vgg16 outperformed the ordinary network by 90.1% for Vgg16 compared to 62.5% for the ordinary convolutional neural network model.\",\"PeriodicalId\":20592,\"journal\":{\"name\":\"PROCEEDINGS OF THE III INTERNATIONAL CONFERENCE ON ADVANCED TECHNOLOGIES IN MATERIALS SCIENCE, MECHANICAL AND AUTOMATION ENGINEERING: MIP: Engineering-III – 2021\",\"volume\":\"60 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-08-19\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"PROCEEDINGS OF THE III INTERNATIONAL CONFERENCE ON ADVANCED TECHNOLOGIES IN MATERIALS SCIENCE, MECHANICAL AND AUTOMATION ENGINEERING: MIP: Engineering-III – 2021\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.58190/icat.2023.20\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"PROCEEDINGS OF THE III INTERNATIONAL CONFERENCE ON ADVANCED TECHNOLOGIES IN MATERIALS SCIENCE, MECHANICAL AND AUTOMATION ENGINEERING: MIP: Engineering-III – 2021","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.58190/icat.2023.20","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Utilizing Transfer Learning on Landscape Image Classification Using the VGG16 Model
In recent times, the need for the use of image classification techniques of machine learning to solve worldly problems in various areas such as agriculture, the health sector, and tourism is rocketing up day by day. Traditionally, one of the most used techniques in image classification is the use of deep neural networks called convolution neural networks (CNN). To come up with a good network model, one needs to have an enormous quantity of data in the form of images and design a network model from scratch in a trial-and-error way. This not only takes a lot of time but also requires very powerful computation equipment such as graphical processing units (GPU). To overcome such barriers, a machine learning technique called transfer learning enables the use of already trained network models in the form of fine-tuning them to solve related issues. In this work, the 2014 ImageNet winner model called Vgg16 was adopted to classify landscape images in the Intel dataset. The dataset contains 5 categories of images namely buildings, forest, glacier, mountain, sea, and street. The performance of Vgg16 was compared to that of a 7-layer ordinary convolution neural network and the results showed that transfer learning with Vgg16 outperformed the ordinary network by 90.1% for Vgg16 compared to 62.5% for the ordinary convolutional neural network model.