Robust Classification of Indian Currency: Deep Learning Solutions for Modern Financial Systems

The paper introduces an innovative model for the classification of Indian currency notes, with five key denominations: Rs 10, Rs 50, Rs 100, Rs 500, and Rs 2000. Exceptional performance metrics have been achieved in the model, with precision, recall, and F1-scores of 94.34, 95.89, 95.42, 94.83, and 95.74 for Rs 10, Rs 50, Rs 100, Rs 500, and Rs 2000, respectively. The dataset had 12,800 images, and the version reached an accuracy of 95.31% while identifying Indian forex notes. These values show that the model is ideal to identify and classify every currency denomination correctly. In addition, the analysis highlighted support and support ratios with a fair distribution of each class; Rs 10 had support of 2120 (17%), Rs 50 at 2920 (23%), Rs 100 at 2620 (20%), Rs 500 at 2320 (18%), and Rs 2000 at 2820 (22%). The attained overall accuracy is as high as 98% for all classes, which is an indication of the reliability of the model in actual operation. Results indicate that the proposed model facilitates high efficiency in overcoming currency detection and verification problems, thereby paving the way toward improvements in automatic financial transaction systems. In this work, advanced classification techniques are taken as a framework for ensuring enhanced security and efficiency in currency management.