Optimasi Klasifikasi Buah Anggur Menggunakan Data Augmentasi dan Convolutional Neural Network

Abstract

Grapes are a popular fruit and can be easily found almost anywhere in the world. Many will be amazed by the sweet and delicious taste of this grape. Grapes not only brings extraordinary delicacy to our taste buds but also bring special benefits to human health. Therefore, the researchers tried to make a grape image recognition program that uses Data Augmentation and Convolutional Neural Network algorithms. It is a convolutional activity that combines some preparatory processing with several components moving together through a biological sensor system. The grapes used are Champagne, Concord, Cotton Candy, Chris Monceedless, Gewürztraminer, Grenora, Kyoho, Moondrops, Pinot Noir, Riesling, Sultana, Sweet Jubilee and Valiant. Classification optimization was carried out on grape images using two test models, namely the sequential model and the on-top VGG16 model, which operate on the Google Collaboratory Website application and Keras. The test data for this observation on training data are 2400 images and test data as many as 480 images that create a value for the sequential model with an accuracy of 98.54% and a loss of 0.027%, for the on top model VGG16 the accuracy value is 99.37% and a loss value is 0.029%