Cowpea leaf disease identification using deep learning

Abstract

The identification of plant/crop diseases is of great interest to agriculture and in turn, the growth of nations. A systematic way of grouping different plant and crop diseases is essential for identifying and cataloguing the extensive information collected about the many different known plant diseases. There has been no significant research on classifying diseases in cowpea plants. This research employs Vision transformer model to classify the diseases found in the Cowpea plant, a native plant found in Kerala. Classifying diseases affecting cowpea leaves is a vital aspect of crop management and protection. Understanding the specific types of diseases present in a crop facilitates early detection and helps implement control measures. Accurate disease classification provides valuable insights into the impact of the disease on the crop. Along with this, for each classified disease, a Possible Treatments and Prevention Methods suggestion is proposed. Six deep learning models—InceptionV3, VGG16, VGG19, Ensemble model, CNN and Vision Transformer (ViT) models were compared for the prediction of diseases in cowpea plants using a dataset of 5100 images of diseases of cowpea plant leaves. ViT outperformed other models with accuracy of 96%. Disease classification helps monitor the disease's spread and evolution, leading to a better understanding of its impact on cowpea production and the development of more effective control methods for researchers, farmers, and agronomists.

Keywords

• Cowpea plants
• Deep learning
• Plant disease classification
• Vision transformer model (ViT)
• Treatment and prevention methods