Deep Learning for Anthracnose Diagnosis in Turnip Leaves

Fungal diseases in plants are extremely pressing issues in the agricultural industry, threatening global food security by reducing crop yields and quality. Traditional approaches to disease diagnosis and management have failed to recognize symptoms when they first appear. Leaves of the turnip, a plant of high agricultural value, has been especially affected by the fungal disease Anthracnose. Therefore, this study aimed to develop a novel convolutional neural network that can identify turnip leaves with early symptoms of Anathrosce blight. The model had 4 convolutional blocks and was trained on a custom dataset of 1,470 images, randomly split into 60% train, 20% validation, and 20% test. To compare how the CNN model fared with other machine learning algorithms, a support vector machine(SVM) model was developed and trained with the same image dataset. The CNN model’s accuracy 98.75% compared to the SVM model’s 80.50% accuracy. These results validate the efficacy of the CNN model to accurately identify infected turnip leaves and demonstrate that it can be implemented into a practical disease diagnosis system. Future studies are warranted to improve the model through means such as k-fold cross validation as well as apply the model architecture to other crops and diseases.