A Comparative Analysis of Multi-Plant Disease Detection and Classification Using Neural Network Architectures

Plant diseases severely impact agriculture and economies, causing significant c rop l osses. E arly a nd accurate detection is essential to minimize damage and spread. Traditional manual inspection methods are time-consuming and error-prone, highlighting the need for automated solutions. This research investigates the use of neural networks for multi-plant disease detection and classification t o p romote s ustainable agriculture. Data from bell peppers, potatoes, tomatoes, rice, and maize were preprocessed through scaling, grayscale conversion, and augmentation. Multiple deep learning models, including CNN, MobileNetV2, VGG16, Hybrid CNN-LSTM, InceptionV3, and EfficientNetB3, w ere e valuated. A mong t hese, EfficientNetB3 consistently outperformed others, achieving notable accuracy rates using an 80-20% data split: 99.59% for bell peppers, 98.04% for potatoes, 92.52% for tomatoes, 97.52% for rice, and 96.25% for maize. For the 70-30% split, EfficientNetB3 achieved 94% for bell peppers and 84% for potatoes, while for the 90-10% split, it maintained competitive performance with 89% for bell peppers and 90% for potatoes. Video classification using EfficientNetB3 f urther d emonstrated i ts e ffectiveness in identifying diseases. These findings u nderscore t he p otential of EfficientNetB3 in enhancing plant disease detection accuracy and efficiency, c ontributing t o s ustainable a gricultural practices.