Evaluating CNN and Traditional Machine Learning Approaches for Early Detection of Potato Infections by Alternaria Solani and Phytophthora Infestans

Potato crops are majorly affected by fungal and oomycete diseases; the most popular and harmful diseases are early blight and late blight caused by Alternaria solani and Phytophthora infestans. In most cases, their symptoms are visually similar in the early stages. It is difficult to accurately identify them and apply specific treatment or implement other disease management strategies. Identification procedures based on Artificial Intelligence, especially the Deep Learning and Machine Learning techniques, are always working effectively compared to the manual or traditional procedures. This study presents the differences between the performance of a deep learning model, such as a CNN, in comparison to a few popular traditional Machine Learning classifiers for identifying Alternaria solani and Phytophthora infestans infections in potato plants. We selected, trained, tested, and validated models of a Convolutional Neural Network (CNN), Support Vector Machine (SVM), and K-Nearest Neighbours (KNN) on an expertly vetted dataset of images of leaves with early blight, late blight, and healthy types. The CNN model trained on some raw image data performed better than the other two models, achieving anaccuracy of 91.7%. On the other hand, the traditional models, such as SVM and KNN, which depend on manually selected features, achieved accuracies of 68.3% and 14.3%. The results powerfully highlight the advantage of CNN-based deep learning models in handling complex tasks such as plant disease classification, owing to their exceptional capability for learning meaningful features. The significant performance difference has been noticed in this work, which strongly confirms that the deep learning model is the best paradigm in fine-grained visual classification in agriculture.