Enhancing Corn Leaf Disease Identification with DenseViT Model

In India, the harvesting of corn had exhibited an everlasting upward trajectory, as evidenced by the fiscal year 2021 when production outweighed 31 million metric tons. On top of that, there was an enormous spike in this quantity, hitting near 35.91 million tonnes for the fiscal year 2022-23. The substantial increase in corn production in India is driven by government efforts to boost ethanol production and demand in the poultry industry. Corn production has encountered a variety of issues, such as Common Rust, Northern Leaf Blight, and Cercospora Leaf Spot infections, which have had an unfavorable impact on socioeconomic situations. Approaches based on deep learning elevate corn production by utilizing image recognition and predictive analytics to detect and control diseases at an early stage. Additionally, it assists in the development of maize varieties that are resistant to diseases, leading to increased yield and sustainability with the limitation for an accurate classification model. This study introduces DenseViT, a deep learning model combining Vision Transformer (ViT) and DenseNet121, a model which can accurate classify corn leaf diseases, trained and evaluated on 4,354 images of 4 types of corn leaves from the PlantVillage dataset. The DenseViT model achieved a remarkable accuracy of 98.85%, with precision, recall, and F1 scores of 98.18%, 98.17%, and 98.17%, respectively. Despite such outstanding outcomes, the model can be explored further on a larger dataset with variety.