Optimizing Deep Learning Approaches for Accurate Plant Leaf Disease Identification and Classification

Abstract

The world's population is expanding at an alarming rate, the yield of cereal crops needs to quadruple this century to prevent shortages at a time when hunger and malnutrition are becoming as unfashionable in politics and public opinion as human slavery. Conventional plant inspection techniques, which are labor-intensive and expert-only, have a significant impact on crops and agriculture's economy. The paper is mainly concerned with the potential of artificial intelligence (AI) to revolutionize plant disease detection. Robots reduce the use of other organic and pesticide products, lessen environmental damage, and lower production costs. In this study, CNN models are compared for their ability to identify plant leaf diseases with a large dataset. At a learning rate of 0.001 and 64 batches, ResNet101 achieved the highest accuracy of 96.26% in the study. EfficientNetB7 also demonstrates an encouraging maximum accuracy of 95.03%. According to the study, improving plant disease diagnosis accuracy through the use of appropriate model architectures, learning rates, and batch sizes is critical for enabling more efficient and sustainable agricultural operations.