Laras Suciningtyas, Raissa Alfatikarani, Muhamad Bagus Fikril Alan, Fajar Makmun Maimunir, Hapsari Peni Agustin Tjahyaningtijas
Indonesia is a country with agriculture and food crops potential. The agricultural production index in 2021 has grown in comparison to 2020 based on these three plant categories, with food crops reaching 92.51, horticultural crops at 121.39, and plantation crops at 161.85. A food crop called the potato (Solanum tuberosum L.) is utilized in Indonesia as a vegetable and is frequently used in processed food, homes, fast food company, and flour and chip company. Attacks from pests and diseases are one of the contributing elements to the current fall in potato crop productivity. The ability to diagnose illnesses more quickly and with greater accuracy has been made possible by technological advancements. One method for detecting potato leaf disease is artificial intelligence (AI). Unbalanced data is another issue, where the relative data sets proportionate class sizes diverge by a significant enough margin. Data processing for imbalanced or oversampled data is necessary to avoid this. A method for combining various data classes to get a balanced data collection is oversampling. This method is known as data pre-processing. Convolutional Neural Network (CNN) method will be used to process the data after it has been balanced. CNN are a type of deep learning algorithm. The CNN approach uses many processing layers to identify and extract data features. More than one convolution layer can be used by CNN to build complicated features from the network's initial simple features. The goal of the study was to use potato leaf image data based on three class, including potato plants with healthy, early blight, and late blight to diagnose illnesses with the greatest accuracy to farmers or agriculture department. The steps in the methodology are choosing the optimal activation, optimizer, and epoch functions to compare, selecting the best potato dataset, pre-processing, splitting the data for training and testing, determining the classification using CNN, and accuracy. The proposed method delivers accuracy on the potato leaf diseases with an accuracy of 98.33% with a 3X3 kernel, ADAM optimizer, activation function ReLu, and epoch 250. © 2023 IEEE.
Universitas Negeri Surabaya, Department of Electrical Engineering, Surabaya, Indonesia
