Infrability Solutions
Infrability Solutions
Features Outputs Workflow Security FAQ
Request a Demo
On-Prem Cheque OCR & Data Extraction (Arabic + English)

Extract cheque fields with confidence scores, securely inside your bank.

An enterprise-grade, on-prem cheque intelligence platform by Infrability Solutions, designed for high-accuracy banking workflows. Designed for high accuracy, field-level confidence, and a practical improvement loop when confidence is low.

Book a Demo See How It Works
On-prem deployment Arabic + English Confidence per field Human review + retrain loop
Sample Output (JSON)
Each field includes a confidence score and extraction source.
/api/v1/extract
{ "cheque_number": "012345", "date": "2026-01-29", "payee_en": "ABC TRADING CO.", "payee_ar": "شركة ايه بي سي للتجارة", "amount_numeric": "1250.00", "amount_words": "One thousand two hundred fifty", "account_number": "****1234", "confidence": { "cheque_number": 0.98, "date": 0.97, "payee_en": 0.92, "payee_ar": 0.90, "amount_numeric": 0.96, "amount_words": 0.89, "account_number": 0.93 } }
Supports batch processing and review queue when confidence is below your threshold.

What this solution delivers

Built specifically for bank cheque formats with practical governance, not a generic OCR demo.

✓

Field-Level Extraction

Extract the exact fields you care about: cheque number, date, payee (AR/EN), amount (words & numeric), and account number.

%

Confidence per Field

Every extracted value includes a confidence score, enabling straight-through processing or review-based workflows.

AR

Arabic + English Ready

Works with bilingual cheques and handwriting variability, while preserving layout awareness for Qatar-region templates.

✓
Template-aware accuracy Region-based extraction using cheque layout templates to reduce false reads and speed up processing.
✓
Human review queue If confidence is low, route the item for validation and capture corrections for improvement.
✓
Continuous improvement loop Corrections become labeled data for retraining and measurable accuracy uplift over time.
✓
API-first integration Simple REST endpoints for core banking, ECM/DMS, and workflow engines.

Outputs you can use immediately

Designed for downstream systems, audits, and operations teams.

✓
Structured JSON + confidence Clean, consistent schema for every cheque, including per-field confidence and masking options.
✓
CSV export for reconciliation Batch runs export results for reconciliation, exception handling, and reporting.
✓
Audit-ready evidence Keep extraction metadata such as template used, processing timestamp, and review actions.
✓
Review UI hooks Plug into your existing validation UI or use a lightweight review screen for corrections.

How it works

A practical workflow that balances automation and control.

01 · Ingest

Upload or batch input

Process scanned images or PDFs from your capture systems with optional pre-processing.

02 · Detect

Template & ROI detection

Identify cheque format and extract regions (ROI) for the required fields.

03 · Extract

OCR + post-processing

Arabic/English OCR with cleaning, validation rules, and confidence scoring per field.

04 · Decide

Auto-approve or review

If confidence is high: straight-through. If low: route to review and capture corrections.

Security & deployment

Built for banking environments where data cannot leave the perimeter.

✓
On-prem deployment Runs fully inside your data center or private cloud with no external dependencies required.
✓
Offline-friendly architecture Models, configs, and inference run locally. Supports controlled update cycles.
✓
Role-based access (optional) Restrict review screens, exports, and admin actions by operations and compliance roles.
✓
Integration-ready REST APIs and clear output schema to plug into core banking and workflow engines.

FAQ

Common questions from banking and operations teams.

Is this a generic OCR tool?

No. It’s optimized for cheque layouts (template/ROI-aware) and focuses on the exact cheque fields required for operations and compliance.

How do you handle low confidence cases?

You set a threshold. Items below the threshold go to a review queue, and the corrections can be captured as labeled data to improve the model over time.

Can it support multiple cheque formats?

Yes. You can onboard multiple templates and tune ROIs for each format. This is especially useful when you have variations across corporate cheque books.

Does it work with Arabic handwriting?

It supports Arabic text and bilingual cheques. For handwriting variability, we combine layout awareness with OCR + post-processing rules and an improvement loop for hard cases.

What are the extracted fields?

Cheque number, date, payee (Arabic and English), amount numeric, amount in words, and account number, plus confidence per field.

Request a demo
Share your cheque formats and target fields, and we’ll propose a fast, controlled POC.
This form is a template. Connect it to your backend or email handler.
Note: We can deploy as a containerized service, VM-based service, or integrate into your existing orchestration.