NetraAadhaar: A Deep Learning-Driven Aadhaar Verification Platform for the Aid of Visually Impaired

Abstract

Visually impaired individuals face challenges in verifying Aadhaar cards due to the absence of braille, often relying on others for assistance. Despite the widespread use of Aadhaar cards across India, research on assistive technologies for such users remains limited, making this work uniquely significant. To address this, we introduce NetraAadhaar, a deep learning-based mobile application designed to assist visually impaired individuals in Aadhaar card verification. The framework consists of: (i) text region extraction from Aadhaar cards using YOLOv8, (ii) recognition of extracted text via Tesseract OCR engine, (iii) text-to-speech conversion for auditory verification, and (iv) the developed framework is fine-tuned and seamlessly integrated into an end-to-end mobile application for real-time use. NetraAadhaar achieves an mAP-50 score of 92.5% for text detection and an overall text recognition accuracy of 87. 79%, with a precision greater than 90% in key classes, demonstrating its robustness. Compared to traditional OCR-based assistive tools, NetraAadhaar offers higher accuracy, real-time performance, and an end-to-end automated pipeline making Aadhaar verification significantly more accessible and reliable for visually impaired users. NetraAadhaar empowers visually impaired users to verify their Aadhaar cards independently, reducing reliance on others and mitigating risks of identity fraud.