Models / moi

moi/anpr-gcc-plates

ANPR GCC Plates

Entity-PrivateHigh riskGov ReuseComputer VisionONNX RuntimeN/A2.2.0

Test in Sandbox

480 19K/30d 12

Plate detection mAP@0.5

0.94

End-to-end plate accuracy

0.89

Character error rate

1.8%

Latency p95 (T4)

38 ms

About this model

Automatic number-plate recognition pipeline covering Qatari and all GCC plate formats, including bilingual Arabic-Latin plates and diplomatic series. Combines a plate detector with a sequence recogniser tuned for gantry and mobile-patrol camera angles.

Intended use

Deployment inside approved MOI traffic-enforcement and border-gate systems only; every downstream use requires an entity-level authorisation record.

Training data lineage

Trained on internally annotated gantry captures with format priors validated against vehicle-movement-aggregates registration distributions.

Limitations & bias notes

Recognition accuracy drops on damaged or retro-reflective plates at night and on motorcycles at oblique angles. Newly introduced plate series require a format-table update before they are recognised reliably.

Evaluation metrics

Plate detection mAP@0.5	0.94
End-to-end plate accuracy	0.89
Character error rate	1.8%
Latency p95 (T4)	38 ms

#anpr#license-plates#gcc#detection#onnx#enforcement

Try it

Live sandbox

gantry_cam_e14_frame_882.jpg

fixture

Evaluation only: gated models run against redacted sample data until your entity's entitlement is approved.

Feedback

Owning entity

moMinistry of Interior

Updated: 2026-03-27
Latest version: 2.2.0
License: Government Internal Reuse
Access: Gated

Trained on

vehicle-movement-aggregates →