Preview ready-to-license photographs and scans with real, readable text in many languages and scripts. Built for OCR, document AI and multimodal teams that want to test image data before commissioning a custom collection.
Infographics
Posters, charts, diagrams and data graphics in many languages and layouts, for document and chart understanding models.
Handwriting
Handwriting from many writers, languages and scripts, on varied paper and in different styles, for handwriting recognition training.
Signage
Large-format advertising and public signage photographed at different distances, angles and times of day, for scene text detection and recognition.
Storefronts
Storefronts and shop signs across markets, with local languages, fonts, hand-painted boards and mixed-script names.
Numbers
Images where digits carry the meaning, such as price tags, meters, displays and room numbers, in different numeral systems.
Multilingual
Text-rich images across Latin, Indic, Arabic, CJK, Cyrillic and Southeast Asian scripts, including bilingual and mixed-script signs.
About This Dataset
An off-the-shelf (OTS) image dataset is a set of photographs and scans that has already been collected and can be licensed for AI training. Our images show real text as it appears in the world: on crumpled paper, curved shop signs, glossy posters and weathered billboards, in many languages and scripts. That is what OCR, scene text recognition and document AI models need to learn from.
Because the data already exists, you can open samples, check the image types and languages, and judge quality before you commit. If you need other languages, scenes or volumes, our custom image data collection service collects them to your brief. You can also browse every OTS dataset, including audio and egocentric video.
Real-world images with readable text in many languages and scripts
Charts, handwriting, signage, storefronts and numeric text
Samples you can preview before you request a dataset
Custom image collection available when OTS data is not enough
Why Real-World Images
Models that read text in images are only as good as the variety of images they trained on.
Images with blur, glare, odd angles and busy backgrounds help OCR and scene text models read text the way it really appears.
Models that must read many languages need examples in each script, including right-to-left and mixed-script text.
Lighting, distance, damage and clutter in real photos are what clean scans and synthetic images miss.
Charts, forms, posters and infographics teach models to interpret layouts, not only isolated words.
Use Cases
These are the most common ways teams use text-rich image data.
Photos of signs, shop fronts and billboards train models to detect and read text in natural scenes.
Pages from many writers, pens and papers train models to read cursive, print and mixed handwriting.
Infographics, charts and mixed text-and-graphic layouts feed document AI and chart-parsing models.
Text-rich images paired with transcriptions and metadata support multimodal training, local search and mapping AI.
Working with other data types? See our audio data collection and text data collection.
Buyer's Checklist
A dataset is only useful if it matches your languages and conditions. Ask about these five things.
Confirm which languages and scripts are included, and whether right-to-left and mixed-script images are part of the set.
Look for different angles, distances, lighting, fonts, handwriting styles and damaged or partly hidden text.
Check whether transcriptions, language tags or text-region labels come with the images, or whether you annotate them yourself.
Know what captured the images, such as phones, cameras or scanners, and whether full-resolution originals are delivered.
Ask how personal details and faces are handled, how contributors gave consent, and what licence terms apply.
Answers to common questions about off-the-shelf image datasets for AI training.
An off-the-shelf (OTS) image dataset is a pre-collected set of photographs and scans that is available to license for AI training. Instead of commissioning a new collection project, you review samples, check that the languages and image types match your model, and start training sooner.
Our image work covers infographics and charts, handwritten notes and documents, billboards and outdoor signage, shop fronts and storefront signs, and images of numbers and numeric text such as price tags, meters and displays. Ask us which categories are available as ready datasets.
Our image collection spans Latin, Devanagari and other Indic scripts, Arabic script, Chinese, Japanese and Korean, Cyrillic, Southeast Asian scripts and mixed-script images. Tell us the languages your model must read and we will confirm what is available.
They can. Depending on the dataset, you can receive language tags, text transcriptions and text-region annotations. Images can also be passed to our annotation services for labeling and review, or delivered as raw images if you annotate in-house.
An OTS dataset is fixed, pre-collected data you can evaluate immediately. Custom image data collection is planned around your own languages, scenes, devices and volume. Many teams start with an OTS dataset to test results and add custom collection for gaps.
Collection guidelines tell contributors what not to capture, and faces and other personal details can be excluded or blurred when a project requires it. Contributors take part with consent, and rights terms are agreed up front. Ask us for the consent and licence terms of the specific dataset you are evaluating.
Typically original full-resolution image files, structured metadata such as language, script, category and capture device, optional labels or transcriptions, and a summary of quality checks. File formats and folder structure are agreed with your team before delivery.
Browse the sample cards above and open any of them to see its description and specification. Then use the contact button on this page and share your image types, languages and the volume you need. We will reply with the matching dataset options.
More OTS Datasets
Need another data type? Browse our egocentric, multi-camera and audio collections, or see all OTS datasets.
Get Started Today
We collect real-world images around your languages, scripts, scenes and quality criteria, with samples to test first.
