NID Data Extraction
End-to-end pipeline that extracts structured fields from National ID card images using OpenCV preprocessing and PaddleOCR.
- Accuracy
- 98%
- Processing
- <2s
- Uptime
- 99.9%
Problem & Solution
The Problem
Manual entry of NID data is slow and error-prone. Organizations need to digitize identity documents at scale with high accuracy across varying image quality.
The Solution
Built a pipeline that preprocesses NID card images (deskew, denoise, contrast enhancement), localizes field regions, runs OCR via PaddleOCR, and outputs structured JSON with name, NID number, date of birth, and address fields.
System Architecture
End-to-end flow from intake to outcome
- 01
Image Ingestion
FastAPIAccepts NID card uploads via API or batch processing queue.
- 02
Preprocessing
OpenCVDeskews, denoises, and normalizes contrast on the input image.
- 03
Field Extraction
PaddleOCRLocalizes and crops field regions, then runs OCR per region.
- 04
Structured Output
PythonValidates and returns extracted fields as typed JSON.
Related Projects
End-to-end cheque digitization pipeline using OpenCV for image preprocessing, PaddleOCR for text and amount extraction, and a Flask + Docker orchestration layer.
- Field accuracy
- 96.2%
- Throughput
- 120 img/s
- p95 latency
- 850ms
Suite of context-aware document intelligence pipelines: PDF intelligence (PyMuPDF + spaCy + BERT), passport MRZ segmentation, and prescription OCR.
- PDF entity F1
- 0.91
- MRZ accuracy
- 98.5%
- Rx field accuracy
- 94%