DocVerify — Photo ID Compliance
AI tool validating and processing government-issued photo ID documents — OCR extraction and compliance checks against formatting rules.
TODO: Add permission-approved, sanitized product screenshots. Concept art is not shown as product evidence.
Technical concept illustration. The case study below describes the project implementation.
ID documents vary widely in layout, print quality, and scan condition, so OCR extraction needed confidence scoring: low-quality or ambiguous extractions get flagged for human review instead of silently passing or failing compliance checks. DocVerify validates government-issued photo ID documents, extracting fields via OCR and checking them against formatting and compliance rules — automating a process that was previously manual document review.
What I Built
- OCR extraction pulling name, ID number, dates, and photo region from scanned documents
- Compliance validation checking format, expiry, and field presence against defined rules
- Confidence scoring flagging low-confidence extractions for manual review rather than auto-approving them
- Audit trail logging every validation decision for accountability
Tech Stack
- Language: Python
- OCR: Tesseract / cloud OCR API
- Computer Vision: OpenCV
- Backend: Flask REST API
Key Decisions
- Added confidence scoring to OCR extraction rather than treating every extraction as pass/fail, so low-quality or ambiguous scans got routed to human review instead of silently passing or failing compliance
- Logged every validation decision to an audit trail, since compliance tooling needs accountability for how each document was assessed, not just a final yes/no output
Outcome / Impact
Automated a manual document-review process, cutting per-document review time significantly while keeping a human-in-the-loop safety net for uncertain cases.
Related work
VisageLab
Computer-vision face-swap research tool built on InsightFace — experimented with realistic face-swapping pipelines for creative applications.
StepPrint
Biometric foot-recognition research prototype — exploring feature extraction and matching for foot-based identification.
GuestGallery
Python app using face recognition to allocate 2000+ wedding photos to guests. Finds each person across photos and surfaces “here are your pictures.” Processes large image sets efficiently.
Interested in similar work?
Review my services and how we can define a useful first milestone.
Discuss similar work