Effective Disease Detection in Bean Leaves Using Deep CNN and SVM Ensemble

Bean leaf diseases have the potential to drastically reduce crop output and quality. The diagnosis of these diseases at an early and accurate stage is critical for optimal care and prevention. In this paper, we propose a model built around a CNN architecture that can forecast the occurrence of six distinct diseases in bean leaves: Acute Leaf Spot, Rust, Bean's Common Mosaic Virus, Bean's Yellow Mosaic Virus, Bacteria Blight, and Anthracnose. Four convolutional layers, 4 maximum-pooling layers, and two completely interconnected layers make up the model. A collection of labeled bean leaf photos was collected and prepared to train and test the model. The convolutional layers enable relevant characteristics and disease-specific patterns to be extracted from the photos. The maximum pooling layers aided in minimizing dimensions and focused on critical information. The completely connected layers performed pattern mapping and prediction at the highest level. The accuracy, recall, F1-score, encouragement, and accuracy measures showed that the model performed well in disease categorization. The study revealed the correct identification and separation of the six disorders, demonstrating the efficacy of the suggested model. The findings emphasize the CNN approach's potential for early disease identification and control in bean crops. This study advances plant pathology by offering an accurate and fast approach to diagnosing beans leaf diseases. The proposed framework can help farmers, agronomists, and scholars make informed disease control and preventive decisions. More study is needed to evaluate the model's effectiveness across various bean types and environmental conditions, which will lead to its usefulness in real-world scenarios.