Maize incurs many diseases, but stalk rot has badly influenced the crop yield. A pathologist, extension worker, or experienced farmer can only identify susceptible stalks to determine the accurate application of fungicide to the crop. It is rigorous for the farmers of developing countries to hire them in time. Moreover, a variation in the views of professionals leads to incorrect findings. In this manuscript, pathologists’ discoveries have become a standard to compare the farmer’s detections with an intelligent-based model. The Convolutional Neural Network (CNN) employment sought to identify the resistant and susceptible stalk against stalk rot. The Maize and Millet Research Institute Yousafwala, Sahiwal, was the chosen field for experimentation. Gathering resistant and vulnerable images from maize germplasm, having local origins, progressed via a smartphone. The CNN architecture’s exploration classified the images into two resistant and susceptible classes. The P value (0.00001) calculated by the Chi-square method for resistant and predisposed groups showed highly significant results. An 83.88% achieved accuracy came from the CNN, while 49.5% of the accuracy resulted from the farmer. Recording recall ratio and precision of 0.766 and 0.896 occurred for resistant, and 0.911 and 0.796 were the recordings for susceptible classes by deep learning technique, respectively. The proposed approach is an influential source of detection of resistant lines against stalk rot disease by minimizing the need for pathologists, extension workers, or experienced farmers. It will help farmers to identify the quantity of fungicide against stalk rot and explore lines for resistant breeding programs.
Extension worker, disease, CNN, deep learning
The proposed model identified the resistant lines against stalk rot more accurately, benefiting breeding programs for improving existing high-yielding varieties.