Recognition of diseases of maize crop using deep learning models

Disease attack on crops is one of the most serious threats to the global food supply chain. A proper, comprehensive and systematic solution is required for the early recognition of diseases and to reduce the overall crop loss. In this regard, deep learning techniques (especially convolutional neural networks (CNNs/ConvNets)) are being successfully applied for automatically recognizing the diseases of crops using digital images. This study proposes a novel 15-layer deep convolutional neural network (CNN) model for recognizing the diseases of maize crop. Around 3852 images of maize crop were collected from the PlantVillage data-repository. This dataset contains leaf images of three diseases viz. gray Leaf Spot (GLS), Common Rust (CR) and Northern Corn Leaf Blight (NCLB) as well as the healthy ones. The proposed model showed significant results for recognizing the unseen diseased images of the maize crop. We also employed a few popular pre-trained networks in the transfer learning approach for training on the maize dataset. We presented the comparative performance analysis between the proposed model and the pre-trained models in the result section of the manuscript. The experimental findings reported that our proposed model showed 3.2% higher prediction performance with 3 x lesser trainable parameters than the best-performing pre-trained network (i.e., DenseNet121). The overall performance analysis reported that the proposed CNN model is very effective in identifying the images of maize diseases and also performs quite better than the popular pre-trained models.