Advancing Plant Disease Identification using Convolutional Neural Networks for Precise Diagnosis in Hibiscus Plants

The effectiveness and accuracy of deep learning models for the precise diagnosis of diseases in Hibiscus plants are studied. Gray mold, mainly caused by Botrytis cinerea, in diseased Hibiscus leaves is identified using 560 images collected to form the dataset. Preprocessing of the obtained dataset is performed through resizing, data augmentation, and noise reduction, which ensures the variability of the dataset. The preprocessed dataset is employed to train four different deep learning architectures: custom CNN, ResNet-50, VGG-16, and VGG-19. The benefits of using each model are that they are able to extract features and learn the pattern of the distribution of these features in the presence of the disease. The results are obtained based on the accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve for comparison purposes. According to the results, the best candidate is VGG-19, which demonstrates accuracy at the level of 98.54%, while the precision of 98.80%, recall of 98.50%, F1 score of 98.65%, and AUC-ROC of 99.00% are also high. The confusion matrices provide details from which the correct classification for VGG-19 of healthy and diseased leaves becomes evident. The real-time implementation of VGG-19 is completely reliable. The applied deep learning models provide numerous benefits in predicting the diseases in plants, and the usage of computational techniques will ensure food security and opportunities for current development.