JaalTaka: A Benchmark Dataset for Detecting Counterfeit Bangladeshi Banknotes
Description
The rapid proliferation of counterfeit currency poses a serious threat to global economic stability by driving inflation, raising prices, and undermining public trust in paper currency. These challenges are particularly severe in cash-dependent economies, where reduced acceptability of currency disrupts transactions and hinders financial inclusion. Although substantial research has focused on banknote recognition and classification, resources dedicated to counterfeit banknote detection remain limited. A key obstacle is the difficulty of collecting counterfeit banknotes due to accessibility and security barriers. To address this gap, we introduce “JaalTaka”, a benchmark dataset of Bangladeshi banknotes containing both genuine and counterfeit banknotes. Genuine banknotes were sourced from three government banks in Bangladesh—Sonali Bank Ltd, Agrani Bank Ltd, and Rupali Bank Ltd, while counterfeit notes were obtained through formal collaboration with the Rapid Action Battalion (RAB), a special security force of Bangladesh, for research purposes. To enhance robustness, the dataset includes banknotes in varied physical conditions such as wear and tear, stains, user markings, graffiti, and repair artifacts (e.g., tape or staple marks). Images were captured using smartphone cameras, providing sufficient resolution to capture subtle security features while ensuring cost-effectiveness, portability, and reproducibility. Dataset Distribution: • Genuine Notes: 802 • Counterfeit Notes: 588 • Total: 1,390 Image Capture The security features of Bangladeshi banknotes are distributed across both sides, with many being highly subtle, making it difficult to distinguish genuine notes from counterfeits using a single image. To address this, six images were captured from different regions of each banknote, focusing on the most significant security features. This makes the dataset particularly valuable for multi-view models, where the final prediction can be derived from the combined analysis of all six views. Value of dataset: • This is the first publicly available dataset that provides a reliable foundation for developing effective Bangladeshi banknote authentication systems. • Our dataset helps create automated systems that can cut down the need for manual inspections in high-volume money transaction environments. It can also be used in machines like ATMs and vending machines to make transactions safer and more reliable. • This dataset not only helps mitigate economic losses from counterfeit currency but also provides a valuable resource for researchers and developers, serving as a benchmark for future studies in banknote authentication. We hope our dataset can serve as a benchmark resource to support counterfeit banknote detection, helping mitigate inflation and reduce price volatility. It aims contribute to restoring public confidence in paper currency.
Files
Institutions
- Daffodil International University
- Independent University
- Chittagong University of Engineering and Technology