TASMU AI

Models / moi

moi/qid-document-ocr

QID Document OCR

Government-RestrictedHigh riskGov ReuseOCRPyTorchBilingualv4.1
Test in Sandbox
1.8K 96K/30d 112

Field extraction accuracy

0.981

QID checksum pass rate

0.997

Character error rate (Arabic)

0.014

Character error rate (Latin)

0.008

Median latency

410 ms / page

About this model

Field-level OCR for Qatari identity documents: QID cards, residence permits, and machine-readable passport zones. Extracts and validates the QID number, Arabic and English names, nationality, and expiry date with per-field confidence, including checksum verification of the QID number. Deployed inside e-services onboarding flows across four entities.

Intended use

Automated data entry and verification in authorized government service workflows with a human confirmation step. Access is gated; use outside approved workflows is prohibited.

Training data lineage

Trained on the QID Specimen Templates set (synthetic and specimen documents only, no live citizen documents) with heavy augmentation for glare, skew, and worn laminate.

Limitations & bias notes

Confidence degrades on documents photographed at angles beyond ~30° or under strong glare; such fields are flagged for manual review. Handwritten legacy documents are out of scope. High-risk classification: every production integration requires an approved DPIA under PDPPL.

Evaluation metrics

Field extraction accuracy0.981
QID checksum pass rate0.997
Character error rate (Arabic)0.014
Character error rate (Latin)0.008
Median latency410 ms / page
#ocr#identity-documents#qid#bilingual#field-extraction#onboarding

Try it

Live sandbox
PHOTO

STATE OF QATAR · دولة قطر

QATAR ID CARD

28935612345

MOHAMMED ABDULLA AL-HAJRI

Specimen

Used in workflows

Retiring this asset would require these chains to be re-pointed first.

Feedback

4.0(1)

Owning entity

moMinistry of Interior
Updated
2026-06-30
Latest version
v4.1
License
Government Internal Reuse
Access
Gated

Trained on

qid-specimen-templates →