Late blight and early blight are serious fungal diseases that significantly impact the quality and quantity of potato crops, posing a threat to food security. Improving crop quality and minimizing losses through initial identification and categorization of late blight (LB) and early blight (EB) is essential for increasing food availability. Machine learning (ML) and deep learning (DL) models have played significant roles in early detection, minimizing losses and improving crop quality. In machine learning, intricate features are manually retrieved from input images, and classification is performed. Conversely, deep learning is an efficient approach that automatically learns features. This paper describes a potato disease detection model developed using deep learning viz. ResNet50, Dense121, Vgg19, Vgg16. The authors applied various preprocessing techniques, including height and width shift, rotation, shear transformation, vertical and horizontal flip, and zooming. An image dataset of size 4072 is used to successfully train and evaluate the model. Of the available dataset, 10% of the dataset was utilized for validation, 80% for training, and the remaining 10% of the data was utilized for evaluating the model’s generalization. The outcome shows that the ResNet50 model surpasses other pre-trained deep learning models such as Vgg19, DenseNet121, and Vgg16 with an accuracy of 98.32%.