Reproduce “Mero Nagarikta: Advanced Nepali Citizenship Data Extractor with Deep Learning-Powered Text Detection and OCR” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sisir Dhakal, Sujan Sigdel, Sandesh Prasad Paudel, Sharad Kumar Ranabhat, Nabin Lamichhane
Researchers developed Mero Nagarikta, a system leveraging YOLOv8 for text detection and optimized PyTesseract for OCR, to automate data extraction from Nepali citizenship cards. The system achieved a mean Average Precision of 99.1% for front cards and 96.1% for back cards, facilitating efficient and accurate document management.
Key insights
- Transfer learning with YOLOv8 proved effective for robust text detection on Nepali citizenship cards, adapting to diverse real-world conditions like varied lighting and physical obstructions.
- A comprehensive data preprocessing pipeline, including augmentation and inference-time image enhancements (e.g., grayscale, blur, edge detection), was crucial for improving image quality and model input.
- Post-processing techniques, such as string matching and dictionary-based corrections using RapidFuzz, were essential for refining OCR output and mitigating common errors in Nepali text extraction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.