Deep learning classification of apple leaf diseases: comparison of neural networks
Year: 2024
Type: article
Abstract: The production of apples is a significant segment of the global agricultural industry, often threatened by diseases and pests. This study investigates the use of convolutional neural networks (CNNs) to classify images of apple tree leaves, distinguishing between healthy leaves and those affected by rust and scab. The objective is to develop an approach for the early detection of these fungal diseases. High-resolution images were collected, considering variations in lighting, angles, and backgrounds. Eighteen pre-trained CNN architectures available in Keras were tested and evaluated using metrics such as accuracy, precision, recall, and F1-score. The EfficientNetV2B2 and DenseNet201 networks showed the best results, with an accuracy of 99%. To enhance classification performance, ensemble techniques were explored, including combining all networks and selecting only the most accurate ones. Although promising, challenges such as computational complexity and the need for real-time processing in practical applications remain. The findings demonstrate the potential of CNNs and ensemble methods in supporting early detection of diseases in apple orchards, providing valuable tools for producers to manage infestations more effectivelys.
Source: Sigmae
Institution: Universidade Federal de Alfenas
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: diamond
OpenAlex ID: W4409804535
