Comparative Evaluation of Deep Learning Models for Efficient and Accurate Paddy Leaf Disease Classification

Agriculture relies heavily on the prompt recognition of plant diseases, as delayed detection often leads to reduced productivity and serious threats to food security. This research presents a comparative evaluation of deep learning models for paddy disease classification using the Paddy Doctor dataset, which contains ~10,400 labeled images across 10 disease categories. The dataset was preprocessed by resizing images to 224×224 pixels, applying normalization, and splitting them into training, validation, and test sets in an 80/10/10 stratified manner. Five models-ResNet50, EfficientNetB4, MobileNetV3, DenseNet121, and ViT-B16 were trained for 8 epochs with a batch size of 16. MobileNetV3 achieved the best overall performance with 96.07% accuracy, 95.22% precision, 95.57% recall, and a 95.34% F1-score. ResNet50 and EfficientNetB4 followed closely with accuracies of 95.39% and 95.11%, respectively, while DenseNet121 (94.53%) and ViT-B16 (92.61%) performed slightly lower. The results highlight MobileNetV3’s effectiveness as a lightweight yet accurate architecture, supporting its potential for scalable deployment in real-world Paddy leaf disease detection systems.