This thesis provides a systematic review of currency recognition technologies, with a focus onglobal progress and its applicability to Bangladesh. Based on the PRISMA protocol, 37 highquality studies were shortlisted from more than 200 papers published between 2020 and 2025.The systematic review groups global progress in recognition, counterfeit detection, hybridmodels, and deployment strategies. Globally, the transition from handcrafted features andtraditional image processing to deep learning models like CNNs, YOLO, and VisionTransformers is apparent. In Bangladesh, progress has been made in lightweight CNNs andpublicly available datasets such as BanglaTaka, NSTU-BDTAKA, and JaalTaka, butchallenges remain in the size of datasets, multimodal fusion, and real-world applications. Themajor gaps in the literature are the absence of benchmarking, very little work on the integrationof explainability tools like SHAP and GradCAM, and very few mobile-friendly systems. Thisthesis proposes a comprehensive framework that integrates recognition and counterfeitdetection, hybrid models, and deployment strategies specifically for Bangladesh.Keywords: Currency Detection, Banknote Recognition, Deep Learning, ComparativeAnalysis, Machine Learning, Bangladesh Currency.
Faraji et al. (2026) studied this question.