Implementation of an Improved CNN Model for Detection and Prevention of Plant Diseases Using Deep Learning in Agriculture

Abstract

Plant growth is a crucial requirement for framers since it provides a pathway for their livelihood. Plant damage and growth are correlated with one another. Healthy crop cultivation is a priority for framers. It has long been a source of worry that plant diseases all around the world annually cause huge crop output losses. To prevent crop damage from pathogen infection during crop development, harvesting, and post-harvest processing of the produce, new technology for prompt plant disease detection must be developed and used. This will increase crop yield and ensure the sustainability of agriculture. among the many methods for identifying plant disease. However, appropriate application of such procedures is highly difficult because it necessitates expensive equipment and demands a deep understanding of technology to determine and resolve unusual errors or lab results. The use of detecting plant disease has been thoroughly illustrated in this study, along with its benefits and drawbacks. For tracking and evaluating the crop variance plant disease based on images identification is one of the most important precision agricultural tasks. Prior to recently, the majority photo processing techniques methods some of which still are making use of what some have referred to as ML architectures. In the fields of image recognition and pattern analysis, the DL network is quickly taking the lead as the standard. However, there are few studies on its use in identifying diseases in plant leaves. Software program tells us the identity of a species of plants, its confidence level, and the treatment that may be used to treat it. Since the suggested technique combines statistical ML and picture processing algorithms, it is substantially less costly and takes less time to predict than other DL based systems. The proposed system's accuracy, which is based on Python, is around 90% up. We used the CNN model to build the potato model where accuracy is 98%. We used the VGG-19 model with the accuracy 99% to build the tomato model . Again, we used the CNN model to build the rice model and the accuracy is 93%.