Image Classification of Corn Leaf Diseases Using CNN Architecture ResNet-50 and Data Augmentation

Corn is a crucial crop in global food production. Despite its ability to thrive in various weather conditions, it is susceptible to diseases, particularly affecting its leaves. Corn leaf diseases have detrimental effects on harvest yields, leading to losses for farmers. This research employs a deep learning approach using Convolutional Neural Network (CNN) ResNet50 architecture and data augmentation to classify images of corn leaf diseases. Factors such as data splitting ratios, learning rates, and model optimization are explored, while data is enriched through augmentation techniques such as rotation, shifting, and flipping. Training data indicates that the use of data augmentation influences the model's accuracy in classifying corn leaf disease images. The RMSProp optimizer provides the highest average validation accuracy in Non-Augmentation training. Meanwhile, evaluation results from the confusion matrix show that applying data augmentation with a 90:10 data split, Nadam optimizer, and learning rate of 0.0001 yields the highest accuracy of up to 94.36%. However, assessment of the model's ability to classify images reveals that configurations utilizing data augmentation, a 90:10 data split, Nadam optimizer, and a learning rate of 0.001 demonstrate optimal performance with precision reaching 94.50%.