Plantain (banana) is among the most significant plantation crops, yet its production has been severely affected by leaf diseases. Traditional manual inspection by specialists is not only time-intensive but also prone to inaccuracies. While earlier research primarily targeted one or two specific leaf diseases or relied on a single convolutional neural network (CNN) architecture, this study introduces a deep learning approach designed to identify eight distinct categories of plantain leaf conditions—seven disease types and one healthy category. Multiple CNN models, including those leveraging transfer learning, were assessed for classification performance. The study compares the effectiveness of EfficientNetB0, ResNet-50, DenseNet- 121, and Inception-V3 in detecting these diseases, with all models fine-tuned on the collected dataset. EfficientNetB0 outperformed the others, reaching a testing accuracy of 98.37%. Detailed evaluation metrics—accuracy, precision, recall, and F1-score—for each class are provided through classification reports and confusion matrices. These findings suggest that modern CNN architectures offer a robust solution for early and accurate detection of plantain leaf diseases.