Automated Plant Disease Diagnosis: Revolutionizing Agricultural Productivity using Convolutional Neural Networks

This is the authors' abstract. We don't add key points for this paper.

Farming is one vital sector in all countries' economic and developmental life worldwide, and agriculture contributes so much to food security and livelihood. However, crop and root diseases significantly constrain agricultural production, compromising yields to about 14% globally annually. Plant diseases may ruin crops, leading to substantial economic losses, low-quality food consumed, and lower yields. Early diagnosis of plant diseases is crucial for crop health and production. Farmers used a laborious and time-consuming traditional way of detecting diseases through eye observation for a long time. Agriculture has evolved into more efficient practices; precision agriculture (PA) is a technology that minimizes farmers' accuracy with various losses. However, PA remains in its developmental stage, particularly with technological progress and the discovery of plant disease detection. The development of automated systems in plant disease diagnosis was one of the promising advancements in modern agriculture based on techniques for image-based information. This study uses a CNN model to identify and classify plant diseases in pictures from the Plant Village Dataset. The best accuracy results that the custom CNN obtained on training, validation, and test data were 93%, 90%, and 90%. These results infer that CNN is a good classifier for classifying plant diseases into two categories.