Moechammad Sarosa, Asfiyatul Badriyah, Rosa Andrie Asmara, Mila Kusuma Wardani, Dimas Firmanda Al Riza, Yunia Mulyani Azis
The early identification of grapevine diseases is an important step in maintaining plant health and increasing yields. Grape vines are agricultural commodities with high economic value but are susceptible to various types of diseases. Disease detection technology using convolutional neural networks (CNN), one of which uses the MobilNet architecture, shows good results. This research aims to demonstrate the performance analysis of MobilNet in detecting diseases in grape leaves. The training process uses a dataset of 1600 images of leaves infected with three diseases: Black-Rot, Esca, leaf blight, and one healthy leaf. MobilNet was selected because of its high ability to classify images with limited computational resources. Thus, this method can help farmers implement preventive measures in a timely manner. MobilNet proved to be efficient and effective in detecting grape leaf diseases, achieving an average validation accuracy of 98% and a testing accuracy of 97%. This model can respond quickly to improve the quality and quantity of the grape harvest and prevent the spread of diseases in grape plants to minimize economic losses. © 2024 IEEE.
Malang State Polytechnic, Department of Electrical Engineering, Malang, Indonesia; Malang State Polytechnic, Department of Teknologi Informasi, Malang, Indonesia; Brawijaya Universty, Department of Biosystems Engineering, Malang, Indonesia; Stie Ekuitas Bandung, Department of Management, Bandung, Indonesia