Search

Travelers who lack extensive experience or those visiting multiple countries in a short timeframe often need help to differentiate and identify the values of mixed foreign coins. This paper introduces a model that employs You Only Look Once (YOLO) object detection to identify various foreign coins within a single image frame and thoroughly evaluate its performance. The model, developed on Google Colab, has been trained using a dataset of 8,100 images and successfully identifies 18 different classes of coins, including the US dollar, European euro, and Chinese yuan. Implementing the model as a mobile-friendly web application allows users to upload images of their coins, and the application will provide identification results along with the option to receive the latest currency exchange information. The performance metrics scores of the model were 0.892, 0.932, and 0.925 for Precision, Recall, and Accuracy, respectively, showing satisfactory results. [ABSTRACT FROM AUTHOR]

We present a vision-based approach to ancient coins' identification. The approach is a two-stage procedure. In the first stage an invariant shape description of the coin edge is computed and matching based on shape is performed. The second stage uses preselection by the first stage in order to refine the matching using local descriptors. Results for different descriptors and coin sides are combined using naive Bayesian fusion. Identification rates on a comprehensive data set of 2400 images of ancient coins are on the order of magnitude of 99%. [ABSTRACT FROM AUTHOR]