Dr. KAUSHIK MURALI

Ms.Bhavna Sulagodu

Abstract

Purpose:Visually impaired in India face difficulty & financial exclusion due to a variety of currency notes.We used design thinking & machine learning (ML) to build a device to identify legal currency.

Methods:Authentication marks on various currency notes were reviewed. AI & ML models were studied to identify denomination,structural damage & authentication marks.A prototype device ‘SCRS’, was built through an iterative design thinking model.

Results:Two cameras were required to ensure critical features were captured.Using combination of existing libraries & convolutional neural networks,identification of denomination was possible.YOLO object detection detected watermarks & ML models helped identify any staining or over-writing.Qualitative validation identified 5 critical authentication marks which were re-learnt by the model with 99% accuracy.

Conclusion:Novel Ai & machine learning based device,SCRS can help with currency identification & authentication for visually impaired.

Full Text