I-NusaPlant Apps : Indonesian Medical Plants Identification Using Convolutional Neural Network With Pre-trained Model MobileNetV2

Abstrak

Indonesia has about 30,000 different kinds of medicinal plants, which is a very large number compared to the total of 40,000 that exist all over the world. In fact, Indonesia, along with other Asian countries like China and India, has one of the highest concentrations of medicinal plants in the world. An expert is required to identify functional medicinal plants, but the number that such experts is quite limited. Convolutional Neural Networks (CNNs) and transfer learning can be very effective tools for identifying Indonesian medical plants. These methods have been shown to be highly accurate at classifying different objects. Transfer learning was used because it can reuse the knowledge gained from previous training, which speeds up the process and improves accuracy. This study used 5,000 images divided into 20 categories. The MobileNetV2 model was used, and it achieved 100% accuracy for all categories in the experiments. The Identification Indonesian Medical Plants method in this study has been implemented in the I-NusaPlant mobile-based application. The app's performance was tested, and it was found to use a maximum of 17% CPU and 197 MB of memory. This app works on all Android versions from 8.0 to 13.