Potatoes are one of the most cultivated crops and are a source of nutrients. They are a global priority as staple foods. For successful potato production, a strong food security system can be developed. Tuber crops are excellent sources of vitamins and minerals. For data processing, two categories of potato leaf disease, early blight and late blight, were considered along with healthy leaves. It is best to identify the disease in the early stage, as this is the best way to ensure successful cultivation of the crop. The total number of images used was 2582. comprised 515 healthy images, 1000 early blight images, and 1067 late blight images. The images were collected from two sources of datasets: one from the Plant Village and one from Mendeley. In this study, we propose a methodology that utilizes several machines learning models, including the K-Nearest Neighbor (KNN), Decision Tree, Naıve Bayes, Random Forest, SVM Linear, and SVM (RBF), along with feature selection. The RESNET 101 model was used for feature selection to improve the predictive performance of the model. The proposed model is shown to be highly accurate. This conclusion is demonstrated by the fact that along with feature selection, the proposed model of RESNET 101 with SVM Linear achieved an accuracy of 99.22 and an AUC accuracy of 99.97. These results also show how critical optimized ML models are for agricultural disease management strategies that are proactive. Crop protection and sustainable farming methods are part of the team’s and promote sustainable agriculture.