Mobile Application for Indian Currency Verification Using Transfer Learning
Laxmi Shaw, Rudra Narayan Sahoo · 2024
Counterfeit notes are one of the biggest concerns in day-today transactions, and negatively affect a country’s economy. Nowadays, due to the use of advanced technologies in the production of counterfeit banknotes, it is almost impossible to detect counterfeit currencies with bare eyes. Thus, there is a need to make use of technologies for the detection of counterfeit currencies. Currently, certain dedicated devices are available for detecting counterfeit currency; however, they are not accessible to the common people. This paper proposes a deep learning-based solution for counterfeit currency detection (CCD). Furthermore, the proposed models integration on mobile devices will provide scalability and thus accessibility to every ordinary man. Therefore, three state-of-the-art pre-trained models, VGG16, InceptionV3, and ResNet50 are introduced. The results of each of the models performances are compared and analyzed. The best-performing model has 99% training accuracy and is recommended for deployment as a mobile application. The obtained result is almost outperforming the current state-of-the-art techniques with an acceptable average test accuracy of 97.3% to check whether the currency is fake or real. The best performing model is securely executed on mobile devices.