Paper Details

Abstract

Plant breeders and pathologists face the challenge of identifying maize strains resistant to stalk rot at an early stage. This task becomes complicated since the disease reduces crop yields, resulting in higher production costs and environmental pollution as a result of the extensive use of fungicides. The current method used by experts to detect resistant strains has limitations. It is based on phenotypic and genotypic procedures highly depending on the expert skills and experience. Trained algorithms have the ability to detect patterns in images and can be used to identify strains resistant to stalk rot at an earlier stage and with higher accuracy. The current research tested the performance of convolutional neural networks in image-based detection of resistant strains using python and TensorFlow framework. It was observed that the proposed model exhibited 81.25% accuracy on phenotypic images. The F1 score, recall ratio and precision of phenotypic images were 83.90%, 92.50% and 75.51% for the resistant class while 78.65%, 70% and 90.32% for the susceptible class respectively. The model showed F1 score, recall ratio and precision of 99.5%, for resistant class of genotypic images. The current study not only helps agriculturists in early identification of the resistant lines against stalk rot in maize but also helps in addressing food security issues worldwide. It also helps in controlling environmental pollution by minimizing the use of fungicides.

To Cite This Article