Deep-Learning-based Technique for Detection and Recognition of Indian Currency Denominations

As human society progressed, trade evolved from the barter system to the use of currency. Each nation has its own set of coins and paper bill that make up its currency. In many financial systems, currency recognition and detection are critical tasks particularly in countries with diverse currency denominations like India. It is a very challenging task to identify fake or real currency for a normal human being as well as a visual impaired person. A deep learning model called YOLO-v8 is suggested as a solution to this problem in order to detect and identify Indian currency in real time. The system is trained on images of various Indian currency denominations (10, 20, 50, 100, 200 and 500), collected under different conditions. Techniques for data augmentation are utilized to enhance the model’s resilience to variations such as orientation, lighting conditions, and signs of wear. These images were then separated into training, validation, and test sets after being annotated using the “LabelImg” tool. The proposed model has been tested on both real-time and experimental datasets, showing impressive results with a detection accuracy of $97.14 \%$. This approach is especially useful for real-world uses, like assisting both normal people and those who are blind or visually handicapped in recognizing currency and detecting fake banknotes.