ShowUI Desktop
ShowUI Desktop is a dataset dedicated to grounding PC user interfaces, including screenshots annotated and enriched via Gpt-4o with function attributes, appearance, and spatial relationships.
7,496 screenshots with annotations, 328 MB, Parquet format
Free, open and permissive use
Description
ShowUI Desktop contains annotated PC screenshots for grounding user interfaces. The annotations, originally from OmniAct, are enriched by Gpt-4o with detailed information about the appearance, spatial relationships, and intended functionality of UI elements.
What is this dataset for?
- Train models to understand and locate items in graphical interfaces
- Improving user interface recognition systems for automation
- Human-computer interaction and AI research for UI grounding
Can it be enriched or improved?
Annotations can be completed or corrected manually. The annotation rewriting code provided allows this dataset to be extended to other screenshots or contexts.
🔎 In summary
🧠 Recommended for
- Researchers in human-computer interaction
- AI UI developers
- R&D vision teams
🔧 Compatible tools
- Hugging Face Datasets
- Frameworks vision
- UI annotation tools
💡 Tip
Use rewriting code to enrich or adapt annotations to your own interfaces.
Frequently Asked Questions
What is the specificity of the ShowUI Desktop dataset?
It focuses on grounding PC interfaces with rich annotations for appearance, function, and spatial relationships.
What is the size of the dataset?
It contains 7,496 annotated screenshots, totaling 328MB of data.
Is the license clear for commercial use?
The license is not explicitly specified, prior checking is recommended before commercial use.
