Agriculture’s significance in sustaining human life has never been more pronounced, with technological integration and digitalization becoming integral to enhancing productivity and ensuring food security. One critical aspect of safeguarding agricultural yields is the early detection and also the classification of various crop diseases. In this study, we hear by present a comprehensive methodology for the detection and classification of diseases affecting potato plants. Leveraging the widely accessible and reputable Plant Village Dataset, encompassing diverse potato disease images, our approach unfolds in three key stages: (1) comparison of various feature extraction methods including pixel feature extraction, LBP feature extraction, and HOG feature extraction, (2) classification of the images employing a range of machine learning models including Multinomial Naive Bayes, Random Forest, Logistic Regression, SVM model, Decision Tree, K neighbors classification, and XGB classifier, and (3) evaluation of the F1 score for each model on each disease class, with results visually depicted through graphs. Our experimental findings reveal that the XGB model has achieved the highest accuracy of 95.3% for Pixel feature extraction, Random Forest outperforms the rest for LBP feature extraction with an accuracy of 80.3%, and HOG feature extraction excels with the SVM classifier with an accuracy of 93.3%. These results guide in choosing the most appropriate techniques for extracting features for machine learning models, maximizing the F1 score. Importantly, the trained models are envisioned to be deployed by farmers for early disease detection and classification, serving as a preventive measure against emerging diseases in potato crops. This research signifies a vital step towards sustainable agriculture, where technology and machine learning empower farmers to protect their yields and ensure food security.