Identifikasi Daun Tanaman Herbal Berkhasiat bagi Kesehatan Menggunakan Arsitektur ResNet50 Berbasis Website

Abstract

Indonesia is known as a country that has tropical forests with biological riches that have more than 30,000 species of plants growing, of which 9,600 species have been identified as having properties that are beneficial to human health. However, there are still errors in identifying these herbal plants, especially the leaves of herbal plants because they have almost the same characteristics, especially for people who do not have any knowledge about these plants. The research was carried out using deep learning methods with the ResNet-50 model to classify herbal plant leaves. This research was carried out by carrying out several stages, including dataset collection, data augmentation, model building, data training, and testing. The research results show that the model used achieved a training accuracy level of 99.74% and validation accuracy of 98.75% at the 30th epoch. Thus, the website-based herbal leaf classification system developed in this research is expected to be a useful tool in supporting understanding of the use of herbal plants for their health benefits.