Sistem Klasifikasi Citra Simplisia Fructus Dalam Obat Tradisional Madura Menggunakan Transfer Learning Pada Algoritma CNN

Abstract

Herbal plants have long been a part of traditional medical practices in various cultures, including in Madurese traditional medicine. One crucial component in the preparation of traditional medicine is Simplisia Fructus. Accurate knowledge of Simplisia Fructus is often a challenge, especially for those unfamiliar with it, as recognizing the various forms of Simplisia Fructus can be difficult due to its numerous types. The utilization of artificial intelligence technology, such as Convolutional Neural Network (CNN), can be a solution to assist in identifying and introducing types of Simplisia Fructus. This research employs transfer learning tested on a small-scale dataset. The dataset comprises six classes: Piperis Nigri Fructus (Black Pepper), Piperis Albi Fructus (White Pepper), Cumini Fructus (Cumin), Amomi Fructus (Cardamom), Tamarindus Indica Fructus Piper Retrofractum Fructus (Javanese Chili), Capsici Frutescentis Fructus (Bird's Eye Chili). The total dataset for all classes is 979. Dataset preprocessing involves dividing it into three parts: 80% for training, 10% for validation, and 10% for testing. Model evaluation using a confusion matrix yielded an accuracy rate of 97%. Additionally, web system testing using blackbox testing resulted in a 99.17% rating in the "Highly Acceptable" category. The system implementation follows the software development life cycle (SDLC), specifically the waterfall model for software development and web coding using the Flask framework. The outcome of this research is a web-based application capable of recognizing types of Simplisia Fructus within the category of Madurese traditional medicine