Efficient CNN for Disease Detection in Wheat for Edge Devices

Accurately and promptly identifying crop diseases is crucial for preventing crop wastage and ensuring high-quality yields. This research paper presents a deep learning-based approach for the real-time detection of diseases in wheat crops using a Convolutional Neural Network (CNN) on edge devices, such as drones. However, deploying traditional CNN models on edge drones is challenging due to their high computational intensity. To overcome these challenges this paper proposes a CNN model that implements parallel feature extraction pipelines which significantly reduces the number of training parameters of the model. In addition to this, a new dataset was also formulated using the LWDCD2020 dataset and different sources from the internet. Transfer learning models RasNet50 and VGG16 were trained to do a comparative study between the newly formed dataset and the LWDCD2020 which showed significant improvement in the new dataset. The result of the proposed model demonstrates that the proposed model outperformed existing state-of-the-art Transfer learning models ResNet50 and VGG16 in terms of computational efficiency and received an F1-score of 0.96 that far exceeds its counterpart. The proposed CNN model using edge devices like drones can help farmers to identify diseased areas promptly and take appropriate measures to prevent the spread of diseases across the field.