A Feature Based Classifier for Bangla Currency Using Deep Learning

As it is the age of automation, this study proposes an innovative approach for the accurate and adaptable detection and classification of Bangladeshi currency notes. Also resolving ongoing issues with traditional approaches, this approach changes the realm of currency detection and classification by easily combining computer vision and machine learning techniques, breaking out the restrictions of standard Convolutional Neural Networks (CNNs). This method ensures quick feature extraction by leveraging the YOLOv8 object detection model, which is enhanced by a ResNet-50 deep learning model for well-constructed classification. A carefully prepared custom dataset containing a wide range of situations and denominations of 10 to 1000 Taka notes ensures that the approach is applicable and useful in the real world. The particular combination of YOLOv8 with ResNet-50 achieves outstanding results, with an amazing 99.83% accuracy in classification. Extensive studies highlight the system's reliability in a wide range of external circumstances, demonstrating its potential for revolutionary currency identification in areas such as economic activity, financial transactions, and more. This particular study marks a new maturity in currency recognition. This marks a significant improvement in accurately detecting and classifying Bangladeshi currency notes, making them more sustainable and effective for modern usage.