Recognition of plant leaf diseases based on deep learning and the chemical reaction optimization algorithm

Abstract

Agriculture plays a crucial role in developing countries such as Vietnam, where 70 percent of the population is employed in agriculture, and 57 percent of the social labor force works in the agricultural sector. Therefore, crop productivity directly affects the lives of many people. One of the primary reasons for reduced crop yields is plant leaf diseases caused by bacteria, fungi, and viruses. Hence, there is a need for a method to help farmers identify leaf diseases early to take appropriate action to protect crops and shift to smart agricultural production. This paper proposes lightweight deep learning (DL) models combined with a support vector machine (SVM), with hyperparameters fine-tuned by chemical reaction optimization (CRO), for detecting plant leaf diseases. The main advantage of the method is the simplicity of the architecture and optimization of the DL model’s hyperparameters, making it easily deployable on low hardware devices. To test the performance of the proposed method, experiments are performed on the PlantVillage dataset using Python. The superiority of the proposed method over the well-known visual geometry group-16 (VGG-16) and MobileNetV2 models is demonstrated by a 10% increase in accuracy prediction and a decrease of 5% and 66% in training time, respectively.