A working Python-based utility for automated Aadhaar detection and masking — deployable today as a baseline, with a clear path to production-grade accuracy.

The current implementation combines PaddleOCR and YOLO to deliver an end-to-end masking pipeline on Linux.
YOLO-based detection of Aadhaar-related regions within scanned documents and images.
PaddleOCR reads and extracts Aadhaar number text from identified regions.
Masking logic is applied on extracted PII, redacting Aadhaar identifiers in the output.
Folder-based, batch-mode processing via Linux cron scheduler — no manual intervention needed.

Performance Baseline
Accuracy is functional but variable — influenced by several real-world document factors.
Low-resolution scans, poor lighting, or compression artefacts reduce OCR confidence significantly.
Skewed, rotated, or busy-background documents affect YOLO detection reliability.
Different Aadhaar layouts, generations, and laminated copies introduce variance in pipeline output.
Deployment Readiness
Important: Current accuracy of 60–70% must be disclosed upfront before any deployment commitment.
Ready Within 1 Week
The existing utility can be stood up rapidly in a Linux environment for pilot or PoC use.
Install packages and configure services.
Provision Linux VM and network.
Configure scheduler and data paths.
Run pilot tests and verify results.
Deploy the baseline utility within a week — and build toward production-grade accuracy on your own data.