Towards Smart Agriculture: Automated Detection and Diagnosis of Bean Leaf Diseases Using ResNext50 Model

Bean leaf halo blight is a widespread plant disease that causes large output losses by affecting the leaves of bean plants. This disease is known as bean leaf halo blight. To develop successful methods for illness management, it is essential to determine the seriousness levels of this disease in a precise and timely manner. The proposed research study present a multi-classification framework that makes use of the ResNeXt50 model to categorize bean leaf halo blight according to one of five distinct seriousness levels. The authors want to achieve the goal of developing a dependable and effective system by utilizing the capabilities of deep learning (DL) and transfer learning (TL). This will allow us to provide farmers with assistance in making well-informed decisions on disease management methods. To make the model more robust, we amped up the data by collecting and processing a collection of bean leaf pictures and by using data augmentation strategies. After being trained and tested, the ResNeXt50 model was found to have an accuracy level of 96.37% overall. Following the investigation of performance parameters for each layer, a confusion matrix was developed to evaluate the classification results produced by the model. In addition, the ResNeXt50 model was judged to be superior to other cutting-edge models under its ability to effectively categorize bean leaf halo blight. This was demonstrated by a comparison with other models. The framework that has been developed has the potential to greatly improve disease control practices, contribute to the enhancement of crop yields, and ensure that agriculture remains sustainable. Future research may concentrate on improving the model and evaluating its performance in real-world settings; this will contribute to the practical implementation of the suggested system.