Detection of diseases in cucumber using deep neural networks

Abstract

The cucumber (Cucumis sativus L.), a globally essential crop, faces severe threats from various foliar diseases. This work explores deep neural networks (AlexNet, Vision Transformer, MobileNet, and U-Net) for the early and accurate detection of these pathologies based on leaf images. We analyzed 4,353 images classified as healthy or diseased through advanced preprocessing and data augmentation techniques. The results highlight Vision Transformer as the most effective architecture, achieving 99% accuracy, surpassing MobileNet with similar performance. Meanwhile, AlexNet and U-Net demonstrated more limited performance. The research underscores the practical applicability of these technologies in intelligent agriculture systems, promoting informed decision-making to reduce economic losses and environmental impact. Furthermore, it emphasizes the importance of integrating these tools into low-cost devices for implementation in rural areas. This approach contributes to the sustainability of cucumber cultivation. It sets a precedent for the efficient management of diseases in modern agriculture.