Hoya Species Plant Identification Based on Leaf and Flower Using Convolutional Neural Network Models.

One of the richest Indonesian flora is Hoya plants. Their flowers have aesthetic appeal and are famous among collectors. It also plays crucial roles in ecosystems as pollinators’ hosts and for traditional medicine. Along with research on Hoya plants in Indonesia, the diversity of Hoya species in Indonesia has begun to increase. This raises a problem for researchers, hobbyists, or even the public to determine the type of Hoya. Thus, they need a Hoya identification system based on leaves and flowers, so that they can easily determine the type of Hoya. This study aims to build a Hoya plant identification model based on leaf and flower images using Convolutional Neural Network (CNN) with two architecture comparisons between VGG16 and DenseNet121. The majority of images were taken in Bogor Botanical Garden, and two collectors’ gardens in Cibinong and Bogor, Indonesia. The total number of images is 412 for Hoya flowers and 701 for Hoya Leaves. Data Augmentation is used to increase the number and variation of datasets. The model built can identify Hoya plants based on the leaf using DenseNet121 with a validation accuracy of 76.09%. Meanwhile, the Hoya identification based on flower images using DenseNet121 has an accuracy of 67.89%.