Image Segmentation Of Ornamental Plants Typical Of South Kalimantan Using The Convolutional Neural Network Method

Abstract

One of the very important processes in the recognition of visually presented objects. Image segmentation is one of the important topics in computer science, especially in the field of digital image processing. The research method used is image segmentation using the Convolutional Neural Network (CNN) method; the results obtained in this study are accurate to the image of plants selected as the sample of this study. The dataset in this study used pictures or objects of ornamental plants, namely Black Orchids, Betel Lurih, and Aglonema Tri-Color.  As for the samples used in this study, namely for these three types of objects, 50 pictures were taken for each object used. By using epochs of 15, researchers have determined to reduce system performance time and by epoch times of 17s, 18s, and 24s. The number of epochs that will be used also affects the time that will be taken by modeling training. Due to the increasing number of epochs, the time that will be required for training will be longer. Then, the accuracy value of the data trained is 0.7667 with a loss value of 0.4039, and the val_loss value is 0.4611 with a val_accuracy of 0.7333. The segmentation results obtained using the convolutional neural network model have a fairly good accuracy level of 0.7667 and a validation accuracy of 0.7333.