Leaf and spike wheat disease detection & classification using an improved deep convolutional architecture
Lakshay Goyal, Chandra Mani Sharma, Anupam Singh, Pradeep Kumar Singh
Abstract
Wheat is the third most harvested and consumed grain in the world. However, a large part of wheat crop becomes spoiled due to diseases. There are over two dozen of wheat diseases that are harmful to the crops. Therefore, the manual diagnosis of these diseases becomes very challenging. Automatic wheat disease classification can be helpful in improving the quantity and quality of the crop yield. Further, it can be a useful mechanism for crop quality assessment, and pricing. Deep learning based image analysis has applications in disease diagnosis and classification. The spike and leaves are the most affected parts of a wheat plant. Majority of diseases can be recognized by the characteristics of these parts. The paper presents a novel wheat disease classification method. A new deep learning model is trained to accurately classify wheat diseases in 10 classes. The proposed method has a high testing accuracy of 97.88%. Furthermore, it gives an improvement of 7.01% and 15.92% for the accuracy metric over the other two popular deep learning models – VGG16 and RESNET50, respectively. Experimental results establish that the proposed method performs better on other parameters such as precision, recall, and f-score.