Comparative Analysis of the Architecture of Convolutional Neural Network Algorithms for Disease Prediction in Paddy Plants

Soil fertility, weather, and temperature are essential factors that support the agricultural sector in Indonesia. However, farmers often experience harvest failures due to pest or disease attacks. One of the reasons for this is the insufficient understanding of the disease and improper handling. To make it easier for farmers to get information and understand the types of diseases in paddy plants, a study was conducted utilizing deep learning technology. The study compared three CNN architectures, namely DenseNet121, ResNet101, and MobileNetV2, to classify the types of diseases in image-based classification paddy plants with several hyperparameter settings. The test was conducted using 2,928 datasets divided into training data, validation data, and testing data with four scenarios. The aim was to identify the most effective CNN architecture for classifying six types of paddy diseases. The final results showed that the MobileNetV2 architecture had the best performance in classifying the six types of paddy diseases among all the scenarios compiled.