Automated Detection of Counterfeit Indian Currency by Xception CNN and Edge Image Capture with ESP32
Shubhankar Pandey,
Meera Singh
This paper explores the application of machine learning (ML) in detecting counterfeit Indian currency
using the Xception deep learning model, which analyzes images to distinguish genuine notes from
fake ones. Out of the entire dataset, 70% fuelled the training phase, 20% guided validation, and the
final 10% was reserved for testing to evaluate the model's performance comprehensively. Performance
assessment relied on indicators like correctness and specificity to measure how well the model did.
The Xception model achieved an impressive 99% accuracy during training but showed limitations in
validation, performing at only 10%. These challenges highlight issues like dataset imbalance and the
need for effective feature extraction to enhance reliability. The findings underline the potential of ML
in aiding banks and law enforcement agencies in identifying counterfeit currency efficiently. However,
the study also identifies key areas for improvement, including addressing data imbalances and refining
the model to improve validation effectiveness. Next-stage exploration will focus on enhancing the
model's robustness, incorporating additional features, and transitioning the model toward real-world
applications.
This research demonstrates the promise of ML in tackling the growing problem of counterfeit currency
and sets a foundation for further advancements in this domain.
