IMAGE
COMMERCIAL
LABELED
Contributors in the 54 eligible countries, where everyday writing uses a non-Latin script, choose the script they are photographing, then shoot any text in it, on signs, packaging, documents or screens, close enough that the characters are sharp; Latin-only and unreadable shots are refused. The script tag is applied at capture by someone who reads the script; Devanagari, Arabic script, Thai, Cyrillic, Han, Bengali and Burmese are all in scope. No transcriptions are included. The non-Latin world is where scene-text models are weakest and training data thinnest: MTVQA covers nine languages, and HierText and TextOCR are Flickr-era and mostly English. Per-script counts are available on request.
Volume
25,000+
Photos
Contributors
1,000+
Regions
Countries using non-Latin scripts (54 countries)
Updated
What’s in the collection and how it’s delivered.
Data modality
Images
Annotated
Labeled (script tag)
Exclusivity
Non-exclusive rights
Capture tool
Phone camera (in app)
Use cases
Multilingual scene-text detection and recognition, script identification, text-centric VQA beyond Latin
Format
JPG / PNG / HEIC
Resolution
Native phone camera resolution (JPG, PNG, HEIC)
Total volume
25,000+ photos (approx. 75 GB)
Regions
Countries using non-Latin scripts (54 countries)
Purity level
QA review on every submission
Where this data comes from and how it’s licensed.
Every file comes straight from the contributor who provided it and is rights-cleared before it’s listed, so your license traces back to the source.
Contributor
Provides the data and uploads it straight to Kled.
Kled
Clears the rights before anything is listed.
Your team
Licenses it under the terms below.
License terms
Non-exclusive commercial training license.
Consent
Each contributor opted in to this use when they uploaded.
Specs reflect the current collection and may change as new data is added. Final terms are set in your license agreement.
