Potato Leaf Disease Detection and Classification Using Deep Learning

Agriculture faces challenges in food crop production and controlling plant diseases, often exacerbated by harmful pesticides. Machine learning and computer vision offer a solution by enabling early disease detection, potentially saving crops. This study focuses on using deep learning models like VGG16, VGG19, ResNet50 and MobileNet to detect Early Blight and Late Blight in potatoes. The PlantVillage Dataset’s potato sub-dataset with categorized images (Healthy, Early Blight and Late Blight) is utilized. Preprocessing techniques like resizing, augmentation, histogram equalization and CLAHE enhance model performance. Transfer learning with pre-trained weights fine-tunes the models, with ResNet50 emerging as the top performer. The study underscores the significance of the Area Under the ROC Curve (AUC-ROC Curve) as an evaluation metric and highlights deep learning’s potential in crop disease detection. Future research may expand to include more diseases and explore advanced machine learning techniques to enhance potato crop disease diagnosis.