App2: Model for apple leaf disease detection based on deep learning

Accurate diagnosis of apple leaf diseases is essential for agricultural productivity in the Lima region, which accounts for 80% of apple production in Peru. Efficient management of these diseases is critical for preserving the health of local crops. This study proposes a hybrid model combining Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs) to diagnose diseases in apple leaves. In this approach, the CNN serves as a feature extractor, while the SVM classifies the extracted features to determine the presence of diseases. This method provides an effective diagnostic tool, contributing to improved disease management in their crops. By combining a CNN with an SVM, high performance is achieved in diagnosing apple leaf diseases. The model was trained and validated using the PlantVillage dataset, with data augmentation and a sample from the Plant Pathology 2021 dataset to increase the diversity of the dataset. Augmentation techniques included rotation, flipping, zoom, translation, and cropping. The proposed model achieved 98.25% in metrics such as accuracy, precision, recall, and F1 Score, outperforming traditional CNN models and other hybrid approaches. These results demonstrate the robustness and effectiveness of the model, establishing it as a promising tool for disease detection in agricultural applications.