Bangladeshi Currency Authentication Checking System Using Convolutional Neural Networks

A technology-driven society has several aspects, where the arrival of digital currency cannot replace paper-based currency yet. The challenge remains to use paper currency because of the intricate adulteration systems. This paper aims to propose suitable models for recognizing the counterfeit currency of the Bangladeshi taka. Bangladesh participates actively in the world economy. Maintaining a favorable reputation among international trade partners and ensuring the legitimacy of its currency is crucial for international trade ties. A surplus of money in circulation that results from counterfeiting can exacerbate inflationary pressures. The identification and elimination of forgery currency contributes to price stability and inflation management. Robust counterfeit money identification systems improve the effectiveness of cash handling procedures in banks, corporations, and governmental organizations, lowering the danger of monetary losses and business interruptions. There are various pre-trained transfer learning models, SVM, and KNN algorithms that are practiced for detecting the authenticity of cash. The suitability of a few models has been analyzed in this work. EfficientNetB7 shows the highest accuracy for detection. The pre-trained models, which have already assigned weights to the imageNet dataset, the last layer of improvisation, and the robust dataset help to determine the strength of an identification system for currency.