Deep transfer learning-based automated detection of blast disease in paddy crop

A major proportion of the loss faced by the agricultural industry originates from the diseases of the crop during cultivation. Paddy crop is one of the dominant crops which provides food to a huge population. In this crop, the losses caused by such diseases vary from 30 to 90% of the yield. Therefore, the automated detection of different diseases in paddy crops seeks the attention of the research community. In this context, the present work proposes a deep transfer learning solution for the automated detection of blast disease of paddy, which is the major cause of its yield reduction. For this purpose, an image dataset of healthy and blast disease-infected leave images of paddy crop has been developed. These images are fed to five convolutional neural network-based deep transfer learning algorithms, viz., LeNet, AlexNet, VGG 16, Inception v1, and Xception models for binary classification. The performance analysis of given algorithms reveals that AlexNet provides better results for binary classification with an average accuracy of 98.7% followed by VGG 16 and LeNet architectures having accuracies of 98.2% and 97.8%. So, this deep transfer learning-based approach may assist in reducing the gap between experts and farmers by providing an automated expert advice platform for the timely detection of diseases in paddy crop.