Abstract

In this research, we present a cutting-edge approach to fortify the robustness of Convolutional Neural Networks (CNNs) for multi-class plant disease detection under varied lighting conditions. Leveraging raw image datasets of plant leaves, this study devises and implements an innovative CNN algorithm, incorporating advanced techniques such as adaptive histogram equalization, generative adversarial networks (GANs) for data augmentation, and transfer learning to enhance classification accuracy. Each stage of the algorithm's development is rigorously compared with established methodologies to substantiate the selection of our techniques. Extensive experimental results, depicted through comprehensive comparative charts, highlight the superior performance of our approach in identifying diseases across rice, sugarcane, wheat, and corn leaves. The findings of this research are poised to make a significant contribution to the field of plant disease detection, offering a dependable and efficient solution for agricultural diagnostics.[1][2][3]