The economic condition and living of many countries depend on agriculture which depends on the production and protection of crops from diseases. Many computerized techniques help to improve the plant disease detection capability accurately at an early stage. This study evaluates the performance of different classifiers on public and private datasets using various neural network architectures, including Alex-Net, Google-Net, Efficient-Net, VGG-16, Vgg-19, ResNet-50, ResNet-101, Shuffle-Net, and Xception-Net. The results show that the neural network architectures perform better than traditional machine learning classifiers, with most achieving perfect accuracy scores of 99.9%. The Q SVM (Quadratic SVM) classifier is the best-performing machine learning classifier, achieving near-perfect accuracy scores across all architectures. However, some architectures, such as ResNet-50 and ResNet-101, perform consistently well across many classifiers, while others, such as Efficient-Net, are not as consistent among the top-performing architectures. Overall, the study suggests that using neural network architectures can lead to excellent classification accuracy for this particular data, but traditional machine learning classifiers quadratic SVM can also perform well, with the decision tree classifier being the weakest performer among the classifiers tested.