International Journal of Science and Technology (IJST)

Potato Leaf Disease Detection Using Feature-Optimized Machine Learning Models

Food security and agricultural output are seriously threatened by potato leaf diseases including Early Blight and Late Blight. Through the use of machine learning (ML) and deep learning (DL) approaches to image processing, this work investigates the automated identification of these disorders. Two models were trained using a preprocessed dataset of potato leaf pictures from the Plant Village repository: a Convolutional Neural Network (CNN) and a K-Nearest Neighbours (KNN) classifier. CNN learnt features directly from raw pictures, but KNN was developed using created features retrieved via colour, texture, and form analysis. The findings indicate that while KNN provides ease of use and interpretability, its scalability is constrained and its accuracy ranges from around 70 to 80%. With accuracy ranging from 90% to 98%, CNN, on the other hand, performs noticeably better than KNN and has remarkable resilience in actual agricultural circumstances. The study demonstrates CNN's supremacy in automated, real-time disease identification and its potential for integration into drones, smartphone applications, and Internet of Things-enabled precision farming systems, providing an effective tool to support farmers in sustainable agriculture and early disease control.