Receipt and price tag OCR
Crumpled receipts, shelf labels and price tags with word boxes and key-value links.
Overview
Photographed rather than scanned, because that is how these documents reach a phone app: folded, curled, shot at an angle under fluorescent light. Fields are linked to their values rather than simply detected.
Specification
- Format
- JPG
- Capture
- Handheld phone
- Annotation
- Word box plus key-value
- Fields
- Merchant, date, total, tax, line items
Sample
Illustrative sample. It shows the structure, fields and metadata of this dataset with invented values, not real recordings or records. A real sample pack is sent on request.
| file | width | height | device | lighting | label_type | label_summary | annotator_id | qa_status | consent_ref |
|---|---|---|---|---|---|---|---|---|---|
| sample_0001.jpg | 1920 | 1080 | Android mid-range phone | daylight | word boxes + transcription | 3 items | ann_0040 | accepted | consent_03100 |
| sample_0002.jpg | 3024 | 4032 | Android budget phone | indoor | word boxes + transcription | 5 items | ann_0043 | accepted | consent_03101 |
| sample_0003.jpg | 1080 | 1920 | iPhone | low light | word boxes + transcription | 2 items | ann_0046 | accepted after review | consent_03102 |
| sample_0004.jpg | 4032 | 3024 | Android mid-range phone | daylight | word boxes + transcription | 7 items | ann_0049 | accepted | consent_03103 |
| sample_0005.jpg | 2560 | 1440 | Laptop microphone | indoor | word boxes + transcription | 4 items | ann_0052 | accepted | consent_03104 |
Quality assurance
- Automated OCR diff
- Human correction loop
- Field completeness audit
More in the catalogue
Full catalogue
Retail product recognition
Supermarket shelves and individual products with tight bounding boxes.
Learn more
Financial document OCR
Invoices, statements and forms with word boxes, key-value links and table structure.
Learn more
Facial expression images
Consented facial capture across regions, ages and lighting, labelled by expression.
Learn moreWant a real sample before you commit?
The sample above shows the structure. We also send a representative slice that includes the hard cases, plus the QA report and licence terms. No call required.