The current world is facing tremendous challenges related to food security, which are likely to increase in the future due to rising population and environmental concerns. Perishable product management is highly complex due to perishability dynamics. The quality and price of the product are the main factors that govern the buying tendency for these products. Thus, the main objective of this work is to develop a deep learning-based image processing model that is both accurate and efficient in order to identify products, like pomegranates, depending on their quality. A unique Convolutional Neural Network (CNN) model is developed for this work. For the same reason, pre-trained models from the ResNet family, including ResNet18, ResNet34, ResNet50, ResNet101, and ResNet152, are also modified using transfer learning approaches. The results of every model are compared with respect to performance metrics, including f1-score, accuracy, precision, recall, and execution time. The base CNN model achieved the highest accuracy of 90.66% and performed best in all other performance criterion. With the help of new age technologies like machine learning, deep learning, etc., real-time non-destructive quality estimation of the perishable products can be done in real time to attach dynamic pricing based on their quality. In the long run, this will increase food security by reducing post-harvest losses through the ability to make supply chain management decisions more quickly and correctly.