Ilnur Saitovich Miftahov, Larisa Yurievna Grudtsyna, I. Myshkina
{"title":"Development of Application for Recognition of Object Groups in the Image","authors":"Ilnur Saitovich Miftahov, Larisa Yurievna Grudtsyna, I. Myshkina","doi":"10.37624/IJERT/13.11.2020.3611-3615","DOIUrl":null,"url":null,"abstract":"In this article, we solve the actual problem of recognizing the object group in the image. The solution to this problem resulted in an application developed on the basis of a neural network for automatic object recognition. We considered the problem of recognition of cars and trucks on images of any size. The developed application creates a neural network, allows it being trained, as well as using the trained network to solve the recognition problem. In this work, we created a convolutional neural network that allows detecting the features of cars under consideration and classifying them. The network architecture was selected in such a way that the result of its work was adequate. When selecting, we considered feed-forward neural networks, since they showed themselves in the best way in solving such problems. To select the structure and build this network, we studied the corresponding theoretical material about the main types of neural networks, as well as about the algorithms for their training. To train this network, we used an error backpropagation algorithm based on the gradient descent method when finding the minimum of the activation function. An important point is tracking of the network training results; to calculate the accuracy and error indicators (in time), the developed application creates the corresponding graphs. To recognize cars on an arbitrary image, it was divided into admissible parts, and then the image was transmitted in parts to the trained network. The network operation result is displayed using the graphical interface of the application. The objects \"passenger car\" and \"truck\" are localized and assigned to the corresponding class.","PeriodicalId":14123,"journal":{"name":"International journal of engineering research and technology","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2020-11-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International journal of engineering research and technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.37624/IJERT/13.11.2020.3611-3615","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
In this article, we solve the actual problem of recognizing the object group in the image. The solution to this problem resulted in an application developed on the basis of a neural network for automatic object recognition. We considered the problem of recognition of cars and trucks on images of any size. The developed application creates a neural network, allows it being trained, as well as using the trained network to solve the recognition problem. In this work, we created a convolutional neural network that allows detecting the features of cars under consideration and classifying them. The network architecture was selected in such a way that the result of its work was adequate. When selecting, we considered feed-forward neural networks, since they showed themselves in the best way in solving such problems. To select the structure and build this network, we studied the corresponding theoretical material about the main types of neural networks, as well as about the algorithms for their training. To train this network, we used an error backpropagation algorithm based on the gradient descent method when finding the minimum of the activation function. An important point is tracking of the network training results; to calculate the accuracy and error indicators (in time), the developed application creates the corresponding graphs. To recognize cars on an arbitrary image, it was divided into admissible parts, and then the image was transmitted in parts to the trained network. The network operation result is displayed using the graphical interface of the application. The objects "passenger car" and "truck" are localized and assigned to the corresponding class.