From Inception to Efficiency: Evaluating State-of-the-Art CNNs for Fake Currency Identification

This paper presents a comprehensive comparison of five state-of-the-art deep learning models for fake currency detection using the Indian Currency Dataset. The performance of Inception V3, ResNet50, Xception and EfficientNet B0 models are evaluated, with the experimental findings indicating that ResNet significantly surpasses the performance of other models, achieving an accuracy of 98.78%, while EfficientNet B0 got 97.82% and Xception got 97.09%. An in-depth analysis of each model's performance is provided, discussing their architectural differences and their impact on currency authentication tasks. This study advances the field of automated banknote verification and provides key insights for integrating advanced neural networks into financial security frameworks.