Bananas, whose demand is very high in the global market, are considered one of the best agricultural export products in the Philippines - a country where agriculture plays a significant role in economic development. However, diseases in bananas have caused significant losses to farmers over the years due to low yields, as it significantly affects the growth and quality of the fruits. To solve the problem, studies have shown that early detection of diseases in bananas is essential for the local farmers to determine a cost-effective control measure to perform which helps reduce the infestation, if not eradicate it. Since image processing has proven to be an effective tool for classification and analysis, it was used as the focus of the study. A total of 3000 images of common banana diseases, divided into training, validation, and testing datasets, and whose symptoms are mostly found on the leaves, were collected, preprocessed, and loaded into the four (4) pre-trained convolutional neural network model architectures namely, VGG19, InceptionV3, ResNet50 and EfficientNet which adopted the same optimization and model parameters. To determine the model with the best performance when used in a test dataset, accuracy results and the confusion matrix and classification report were utilized as performance evaluation metrics. The results have shown that among the identified model architectures, the EfficientNet model obtained the highest accuracy of 91%.