Implementasi Algoritma Convolutional Neural Network (CNN) Untuk Identifikasi Jenis Tanaman Rimpang (Zingiberaceae)

Abstract

Rhizomes (Zingiberaceae) are modified plant stems that grow horizontally beneath the soil surface and can produce shoots and new roots from their nodes. Rhizome plants (Zingiberaceae) are known as ginger or spice plants. This research article discusses the identification of rhizome plant species using Convolutional Neural Network (CNN) algorithm with VGG19 architecture, involving a total of 10 classes of data samples. The rhizome images underwent data preprocessing, resizing them from 500 x 500 to 200 x 200 pixels. During the model design phase, three different scenarios were tested, considering variations in dataset proportions, number of epochs, and batch sizes. The results of the three scenarios showed that the second scenario performed the best, achieving an accuracy of 90%, a loss of 0.285, precision of 93%, recall of 89%, and F1-Score of 91%. The first scenario obtained an accuracy of 88%, and the third scenario achieved an accuracy of 82%. However, when applying the model to test images and achieving the highest accuracy of 90% during training, the accuracy dropped to 40% when evaluated on 100 testing data. This drop in accuracy can be attributed to several factors, including noise in the dataset used and insufficient amount of training data, leading to the model being less effective in learning and recognizing data patterns.