Banknote classification is a fundamental component of financial system architecture. Recognition as well as verifying the integrity of banknotes is extremely important for secure transactions at financial institutions. A highly precise multi-classification model can be created using deep learning along with an extensive image dataset of these currency notes featuring them in a myriad of conditions. In addition to providing accuracy, this would also assist reduce the amount of manual intervention required, which would result in greater employee efficiency. This study involves the implementation of transfer learning on a publicly available image dataset of Indian and Thai banknotes. The results of the study will help create an automatic banknote identification and integrity verification layer in the transaction system architecture of financial institutions.