Leveraging Deep Learning for Efficient Bean Leaf Disease Classification in Uganda

Beans are one of the most important plants worldwide, both in their dried and fresh forms. They are a great source of protein and have many health benefits. However, there are many diseases associated with beans that hinder their production. In recent years, plant leaf diseases have become a widespread problem for which requiring immediate and accurate, knowing the actual type of the bean disease is a crucial step in solving the disease problem. Thus, a need for a precise classification approach for bean diseases. We propose a ResNet model, evaluated on a large (7701) collection of public bean leaf images to efficiently categorize bean leaf diseases. New Method: Using a ResNet model with the open-source TensorFlow framework and a public collection of leaf images, an efficient strategy is proposed, for not only identifying infected bean leaves but also categorizing the bean diseases. In this paper, we clearly detail the steps that were undertaken to solve the problem and we explain the importance of each step. In addition, we compared the outcomes of applying each architecture independently in order to determine which architecture configuration produced the best results for bean leaf disease classification. Additionally, an optimization method was applied to emphasize the differences in ResNet model performance. Preprint submitted to Journal of Neuroscience Methods August 27, 2025. Results: Based on the evaluation results, the proposed ResNet model yielded a 11.38% increment in the classification accuracy as compared to the best baseline model. Conclusion: A ResNet model is proposed for identification and classification of bean leaf diseases. This ensures timely intervention thus minimizing losses due to crop diseases.