Indian Coin Detection and Recognition Using Deep Learning Algorithm
S. Prabu, K. Joseph Abraham Sundar, G. Shanmukhanjali, N. Jawali, K. Sharvani, V. Nirmala
Abstract
Coins play a very important role in our daily transactions whereas manually counting them is a bit of a tiring task. There is a need for automatic coin detection and recognition systems. The main objective of this paper is to detect and recognize Indian coins of various denominations based on their features including texture, colour, shape and so on. A very popular object detection algorithm, YOLOv5 is employed to perform this object detection task. CSPDarkNet and PANet in YOLOv5 are used to extract rich informative features and solve the memory and time requirement problem. YOLO layer in the YOLOv5 model generates a vector that includes anchor boxes, class labels and class scores. We demonstrated the performance of YOLOv5 with other popular objects such as Fast-RCNN, Faster-RCNN and SSD. However, the YOLOv5 model comprehensively outperforms in terms of accuracy and speed. Additionally, it also compared with current state-of-the-art methods still the YOLOv5 surpass those methods.