AI-Based Crop Disease Detection: Evaluation of Wheat Rust Disease Detection and Classification Using Deep Learning and Machine Learning Approaches

Wheat rust diseases have a devastating negative impact on the yield of wheat all over the world. Computer vision can play a vital role in their early detection and the mitigation of their consequences. In this paper, the authors have used the Support Vector Machine (SVM) algorithm, Convolutional Neural Network (CNN), and CNN combined with Vision Transformers (CNN&ViT) to analyze and classify plant images from the Wheat Rust Classification Dataset. The results show that CNN&ViT attained the highest classification accuracy of 98.3% while CNN achieved an 95.97% accuracy. In the authors’ experiment, it was further shown that CNN exhibits a faster convergence while CNN&ViT can still improve when increasing the number of epochs. SVM attained an accuracy of 91.7% which compares relatively well with its Deep Learning counterparts. Other performance metrics like precision, recall, and F1-score also returned results that are higher than 90% for each of the three models thus making them suitable for detecting and classifying small to medium-sized image datasets.