MultiUI Dataset
Massive dataset of 7.3 million examples from 1 million websites, combining texts, images, and UI layouts for visual understanding and interaction.
7.3 million multimodal examples (text, images, UI), structured format
Open Data Commons Attribution (ODC-BY)
Description
MultiUI is a very large-scale dataset combining textual data, images and UI elements extracted from web pages. It contains 7.3 million examples from one million sites. This corpus is designed to train models that can interpret complex interfaces, perform OCR, understand documents, and analyze graphics.
What is this dataset for?
- Train models for the multimodal understanding of web user interfaces
- Improving text recognition and document analysis in a variety of contexts
- Develop AI agents that can interact with complex UIs
Can it be enriched or improved?
Yes, this dataset can be supplemented by specific annotations (user actions, errors, types of UI elements) and by data from other types of interfaces.
🔎 In summary
🧠 Recommended for
- UI researchers
- OCR developers
- AI web agent projects
🔧 Compatible tools
- PyTorch
- TensorFlow
- Detectron2
- OCR libraries
- Multimodal frameworks
💡 Tip
Use suitable sub-samples for testing prior to large-scale exploitation.
Frequently Asked Questions
What is the size and diversity of data in MultiUI?
7.3 million multimodal examples from one million websites, including text, images, and varied UI layouts.
What license governs the use of this dataset?
Open Data Commons Attribution (ODC-BY), free license with mandatory attribution.
Is this dataset suitable for OCR and document comprehension projects?
Yes, MultiUI excels in text recognition tasks and multimodal analysis of documents and web interfaces.
