Deep learning, a subset of machine learning, employs artificial neural networks to extract intricate patterns and correlations from data. Its applications span across tasks like image recognition and categorization. One practical use of deep learning is in evaluating the ripeness of bananas to ascertain their shelf-life. The duration for which a banana remains edible is influenced by its level of ripeness, which can be determined through an objective method such as physico-chemical, textural, and optical characterization or subjective approach, mainly the visual observation of the peel's colour and texture. However, manual inspection of banana ripeness is time-consuming, labour-intensive, and subjective. Therefore, an automatic and objective method for banana ripeness classification is desirable for the fruit industry and consumers. In this study, a deep learning model was used to classify the shelf-life of banana based on ripening, using convolutional neural networks (CNNs) and transfer learning. The datasets of banana images used included raw to ripened banana, to train and test our models. We compare the performance of different CNN architectures, such as ResNet-50, and VGG. The results showed that the developed model achieved high accuracy. Furthermore, the present study evaluated the performance of shelf-life prediction models. We also discuss the challenges and limitations of our approach and suggest some future directions for improvement and applications in real-time grading of banana.