<![CDATA[There are about 15,000 to 20,000 orchid species around the world, spread across more than 900 genera. They come in many different shapes, sizes, and colors. This wide range of species makes it hard to tell them apart, especially for people who aren't experts. Bengkulu is one of the provinces on the island of Sumatra. It is known for its historical and cultural heritage as well as its rich biodiversity, especially its native plants like orchids. However, it is still hard to tell what they are. The identification process can be made better by using the breakthrough in artificial intelligence of the transformer. The goal of this study is to create an Android app that can use the Vision Transformer (ViT) architecture to identify five types of orchids: Bulbophyllum, Cymbidium, Dendrobium, Phalaenopsis, and Vanda. We used open-source libraries to collect data, which included 1,500 images that went through preprocessing steps. The experimental results show that the ViT-Base16 model with 25 epochs did the best job, getting an accuracy of 0.98 on the test dataset. However, it was hard to classify the genus Dendrobium in all trials because it had a lot of different shapes. The application testing gave good results, with scores of 81.13 for ease of use, 82.5 for accuracy, and 83.06 for usefulness. These results indicate that the application successfully aids in the identification of orchid genera, serving as a useful resource for both educational and practical applications]]>
Copyrights © 2025
