Dataset · ServiceNow
Interface screenshots and annotations for AI agents
ServiceNow/GroundCUA
A large-scale corpus of screenshots from 87 applications with structured annotations for training computer-controlling agents.
What's inside
Ranging from 1M to 10M records, this large-scale dataset includes real English interface screenshots with detailed element annotations. Collected from human interactions, it covers 87 software platforms: office suites, browsers, development environments, messengers, and system utilities. Created by ServiceNow for graphical user interface (GUI) research.
What you can do with it
- Training multimodal AI agents for navigation and clicking in user interfaces
- Training control element recognition models on screenshots
- Testing desktop application automation systems
- Tasks
- Image → text
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 34.2K
- Likes
- 46
Models trained on it
- sentimentalk/pixeltree-detectorObject detection 90
- adriandj3/TernaryPilot-v1-Computer-Browser-Use-LoRAImage text → text 88
- zcyyyds/Qwen3.5-4B-GUI-EdgeImage text → text 41
- swaynehales/uitag-yolo11s-ui-detect-v1Object detection 33
- kageroumado/rocuronium-ui-detectorObject detection 15
- laywens/uitag-yolo11s-ui-detect-v1Object detection 0
- ComputerUseAgents/yolo-ui-homework-experimentalObject detection 0
- elidot007/S1-Fast 0
How to load it
from datasets import load_dataset
ds = load_dataset("ServiceNow/GroundCUA")