A labelled image dataset of genuine and counterfeit Ugandan banknotes

This data article describes a labelled image dataset of genuine and counterfeit Ugandan banknotes prepared for research on visual currency authentication and machine-learning-based counterfeit detection. The dataset was compiled from smartphone photographs of banknotes in the denominations of UGX 1000, 2000, 5000, 10,000, 20,000, and 50,000, and curated into a single corpus labelled as genuine or counterfeit. The banknotes, especially counterfeit ones, were obtained from mobile money agent shops, micro-finance institutions, banks, and local markets in Uganda. Exact duplicate image files were identified using SHA-256 content hashes. Redundant copies and files that could not be decoded after copying or conversion were excluded from the clean corpus. The clean dataset contains 1426 image files, comprising 666 genuine and 760 counterfeit banknote images. This article describes the master image corpus, organized into genuine and counterfeit banknotes, and its denomination distribution. This pilot-scale dataset can be used for exploratory research, prototyping, and teaching image classification methods for banknote authenticity screening.'