CoinRec: Object Detection Model for Indian Coin Recognition using YOLO

This Traditionally Indian coin recognition and segregation is performed manually. But it is very much time consuming and tedious process. Some researchers used machine vision algorithm to classify different coins. But currently few explanatory analysis of the trained models existed for the above classification. This work mainly focuses on building a machine vision model for the detection and identification of Indian coin denominations using modern deep learning algorithm YOLO. Also Eigen class activation map is used for better visual explanation of the model for the detection of Indian coins. The class activation maps provide a better visual explanation of how the model predicts the denomination of the coin from its image. The performance of resultant models after training and validation is almost more than 93% for the precision and recall. Similarly mAP50 and mAP50-95 of the resultant model is more than 98%.