Identifikasi Varietas Mangga (Mangifera indica L.) Berdasarkan Citra Daun Menggunakan Neural Network

Identification of Mango Variety (Mangifera indica L.) Based on Leaf Image Using Neural Network

Abstract

The large diversity of mango fruit varieties and their distribution is not in line with the completeness and ease of access to related information. Currently, the categorization and identification of plant variants is one of the most significant research problems in the agricultural domain. The focus of this research lies on identifying mango varieties based on leaf shape, because the leaves on each plant are different and can deceive the human eye, causing misclassification of plants. So an objective automation system is needed in classifying mango leaf types such as the backpropagation classification method and CNN. The results showed that the accuracy of the backpropagation method system for 225 datasets reached 74.67%, while the accuracy of the CNN method reached 91.11% (VGG-16) and 93.33% (MobileNet). However, there is an increase in accuracy when data augmentation is performed, so that it becomes 675 datasets. The result of VGG-16 accuracy becomes 96.30% and MobileNet accuracy becomes 95.56%.