Identification of Hoya as a Medicinal Plant Using Convolutional Neural Network

Medicinal plants have been utilized since ancient times as a treatment and prevention of disease. Some of Hoya species possess medicinal properties and has been used traditionally by some tribes of Indonesia and other South East Asian as well as in Asia Pacific region. Extraction from Hoya leaves provides therapeutic properties and a healing effects. This study focuses on identifying and distinguishing various medicinal Hoya species from each other, recognizing that all species in the dataset have known medicinal properties. However, there are some similarities in leaf shape between Hoya species. The purpose of this research is to design an identification system specifically for Hoya species with medicinal properties using a Convolutional Neural Network (CNN). This study aims to build a CNN model by comparing the VGG16 (Visual Geometry Group 16) and MobileNetV2 architectures. The dataset we used consists of Hoya species that are known for their bioactive properties and medicinal benefits. Hoya plant images were obtained from the RIN Dataverse, BRIN, comprising five different types of Hoya, including Hoya diversifolia, Hoya finlaysonii, Hoya imbricata, Hoya latifolia, and Hoya verticillata. The total number of images from these five species is 71. The designed model is able to identify medicinal Hoya plants based on leaf images, achieving a testing accuracy of 80% using the MobileNetV2 and VGG16 reached 53.33%.