Development of an SDG-driven Convolutional Neural Network (CNN) model for multi-class classification of Tomato Leaf diseases using combined public and local datasets

Abstract

Tomato plant is one of the most consumed vegetables globally. Tomato production has a great role in food security that secure the United Nations Sustainable Development Goals (SDGs). Early detection and classification of tomato leaf diseases is vital to mitigate yield loss. This study focuses on the development of SDG-driven Convolutional Neural Network model for the accurate multi-class classification of tomato leaf diseases using a combination of locally prepared and publicly available datasets. The combined data set is labeled with different classes such as: Bacterial Spot, Early Blight, Sectorial Leaf spot, Leaf mold, Yellow Leaf curl and healthy. The study developed custom made CNN, and VGG16 models applying various augmentation techniques, data splitting methods and hyperparameter tuning. The experimental results indicate that the custom-made CNN model outperform the pertained models with an accuracy of 99.8% using RGB image, augmentation techniques, 200 epochs, Adam, 0.0001 learning rate and a testing data set of 15% ratio. In conclusion, the study showed the use of AI-driven method to reduce the dependency on manual disease detection and improve the tomato production.