Convolutional Neural Networks based Classification and Detection of Plant Disease

Abstract

Plant disease detection is vital in agriculture since it is essential for increasing crop yield. Following recent improvements in image processing, visual plant disease analysis is a current tool for dealing with this problem. The problem of plant disease detection, which is done visually for identification of plant disease, is examined in this paper. Compared to other forms of photographic photos, plant disease images are more likely to feature randomly spread lesions, variable symptoms, and complicated backdrops, making discriminative information difficult to capture. PlantVillage Dataset, which contains 20,000 photos divided into 15 classifications, is used to help with plant disease recognition study. The PlantVillage dataset was used to train the models. In this work, the performance of ResNet-50 deep learning architecture was compared against CNN, and deep learning models in the categorization of plant leaf diseases. The PlantVillage dataset was used to train models. This study assessed the performance of ResNet-50 deep learning architecture to CNN, and deep learning models in plant leaf disease classification. When compared to other deep learning models in the original dataset, the test dataset revealed that ResNet-50 architecture had the highest accuracy of 96.63 percent.