Skip to content

Dataset · abdoelsayed

20,000 annotated Arabic and English receipts

abdoelsayed/CORU

Bilingual corpus of 20,000 annotated receipts for training OCR and structured data extraction systems.

What's inside

ReceiptSense dataset includes 20,000 annotated Arabic and English receipts (metadata specifies 10K to 100K records range). Data collected for complex bilingual text recognition and information extraction from financial documents.

What you can do with it

  • Training OCR models to recognize complex Arabic script mixed with Latin characters
  • Developing receipt parsers and object detection field systems for fintech
Tasks
Object detection, Text classification, Zero shot classification
Languages
English, Arabic
Size
10K–100K rows
Downloads this month
4.6K
Likes
20

How to load it

from datasets import load_dataset

ds = load_dataset("abdoelsayed/CORU")