Abstract
Plant disease classification is an important task to increase agricultural sustainability and productivity. This study conducted a comparative analysis of deep neural networks DenseNet, ResNet, and GoogleNet for automatic classification of plant diseases. Data from 19 disease classes including Apple Black Rot, Potato Early Blight, and Tomato Late Blight were used in the study. DenseNet achieved the highest accuracy (99.80%) on the test data, outperforming ResNet (98.95%) and GoogleNet (98.30%) for efficiently extracting complex features. Data augmentation techniques improved the robustness of the models to changing research conditions. The results highlight the potential of DenseNet for diagnostic automation and integration into mobile and drone systems. In the future, it is planned to expand the dataset and use adaptation methods to various conditions.