Classifying five distinct mango species has been accomplished with the implementation of transfer learning (TL) approach. Mango classification is significant because it helps upcoming generations by enabling the recognition of mango species without any prior knowledge. We observe from background studies that there is a severe scarcity of proper mango species paperwork and datasets, especially in Bangladesh. To resolve the mango identification problem, we collected a dataset on the specific number of images per class which includes 1000 images and the mango varieties are Amrapali, Banana Mango, Harivanga, Himsagar, and Langra. Then the dataset was prepossessed by applying techniques such as image resizing, noise reduction. Moreover, A TL approach is used to solve our problem due to the limitations of computational energy and faster execution. To classify the mango species, we applied MobileNetV2, InceptionV3, and Xception models because of their real-time analysis or frequent species classification more than others. Out of all the experiments, our proposed model "MobileNetV2"has outperformed all other models in terms of efficiency, being able to recognize fruit species with an accuracy score of 98.40% and a loss of 4.68%, and this performance was validated by our test set. The collected dataset tends to be beneficial in terms of future research in classifying mango species of the local region of Bangladesh. © 2023 IEEE.