Classification of Wheat Diseases from Images Utilising Convolutional Neural Networks

Wheat yield preservation is a vital goal in agricultural production, with effective disease detection and management playing a crucial role. Recent advances in computer vision and deep learning have facilitated automated and precise identification of plant diseases. This study proposes a Convolutional Neural Network (CNN)-based method for classifying wheat diseases from images. To improve feature extraction, a position attention mechanism is integrated into the network, allowing the model to focus on relevant regions and better capture spatial relationships within feature maps. Additionally, transfer learning is utilized to expedite the training process and enhance performance, particularly when working with limited data. Among the evaluated architectures, a ResNet model combined with the proposed attention mechanism achieved outstanding results, attaining a testing accuracy of 96.4%. Furthermore, testing on an open-source dataset confirms the model’s robustness and its ability to generalize across various disease categories.