HERBAL LEAF CLASSIFICATION USING DEEP LEARNING MODEL EFFICIENTNETV2B0

Abstract

Science regarding plants has experienced significant progress, especially in the study of medicinal plants. Medicinal plants have been used in medicine and are still an important component in the world of health today. Among the various parts of the plant, the leaves are also one that can be used as medicine. However, not many people can recognize these herbal leaves directly. This is because the herbal leaves at first glance look almost the same, so it is difficult to differentiate them. The aim of this research is to classify herbal leaf images by identifying the structural features of the leaf images. The dataset in this study uses 10 classes of leaf images, namely, starfruit, guava, lime, basil, aloe vera, jackfruit, pandan, papaya, celery, and betel, where each class uses 350 images with a total of 3500 images of data. The EfficientNetV2B0 model was chosen because it has a minimalist architecture but has high effectiveness. Based on the results of research using the EffiecientNetV2B0 model, the accuracy was 99.14% and the loss value was 1.95% using test data.