DefinePK

Abstract

Food security and the economy around the globe depend on crops. However, the increasing rates of plant diseases pose a threat to food production and can lead to financial difficulties. Methods for manually identifying diseases are labour-intensive; introductory tests can be inaccurate and are not particularly helpful. To address these issues, this study employs an automated approach utilizing the VGG16 and EfficientNetV2B0 deep learning models on data collected for rice, corn, wheat, and potatoes. Due to resizing, normalization, and the use of rotation, flipping, and contrast-adjusting methods, the models became stronger and can be generalized more effectively. EfficientNetV2B0 performed better than VGG16, with accuracies of 98.32% for rice, 96.40% for corn, 94.86% for potato, and 91.02% for wheat. The models rely on CNN architecture to identify features in leaf photos that support the detection of differences between healthy and ill specimens. ImageNet was utilized to avoid the development of architecture from scratch. By utilizing AI for disease detection, farmers can rely less on manual identification, respond more quickly, reduce pesticide use, and minimize crop losses. The proposed system aligns with sustainable farming by being able to scale to meet the needs of various precision farming methods. Qualities such as shared symptoms and multiple infections on leaves further emphasize the need for new AI techniques, along with a deep learning framework, to explain their results. The study also suggests that future researchers should implement advanced edge computing methods, utilize automated systems powered by IoT, and expand the datasets to capture rare diseases. The study also highlights how integrating advanced computing techniques with agricultural needs can significantly enhance crop yield, manage resources efficiently, and effectively protect the global food supply.