Development of a Classification Model for Banana Leaf Disease Using Google Teachable Machine

This research developed a machine learning model using Google Teachable Machine and the MobileNetV2 architecture to classify various banana leaf diseases. Banana plants, critical to global agriculture, are highly vulnerable to diseases that can greatly reduce both yield and quality. Early and precise detection of these diseases is crucial for effective management and maintaining crop health. The study utilized 408 images of banana leaves showing conditions like Black Sigatoka, Bract Mosaic, Insect Pest, Moko Disease, Panama Disease, Yellow Sigatoka, and healthy leaves. Through image augmentation, the dataset was expanded to 3,264 images to improve the model's ability to generalize. The data was split into 85% for training and 15% for testing. After training for 50 epochs, the model achieved an accuracy of 98.17%, with precision, recall, and Fl-scores all surpassing 97%. The model performed particularly well in detecting Bract Mosaic, Moko Disease, and Panama Disease, while minor misclassifications occurred between visually similar diseases such as Black and Yellow Sigatoka. When compared to other models like AlexNet and LeNet, which reported accuracies of 87.5% and 99.72^%, respectively, the MobileNetV2 model strikes a balance between high accuracy and efficiency. This model shows promise for deployment as a real-time disease detection tool through mobile applications, particularly benefiting small-scale farmers. Future improvements could involve expanding the dataset and optimizing the model for a wider range of environmental conditions to enhance its agricultural use.