Universal Data Tool
An open-source tool and library for creating and labeling datasets of images, audio, text, documents and video in an open data format
New! Manage label quality and your team with UDT SCALE EDITION!
Simplicity and Power
Easy for everyone
The Universal Data Tool can be used by anyone on your team, no data or programming skills needed. Simplicity without sacrificing any powerful developer features and integrations.
Easy/No Install
Use the Universal Data Tool directly from a web browser or with a Windows, Mac or Linux desktop application.
Real-time Collaboration
Join a link to a collaborative session and see dataset samples from team members complete in real time.
Easy Import from AWS
Import from your S3 buckets easily with IAM or Cognito authentication.
Community
Built Community-First
Working together, we can accomplish more. The Universal Data Tool was built to bring together the best ideas from different machine learning communities. Check out our Slack, Contribution Guide and Community Update Video Series.
Training
Train your labelers with Courses
Upload your dataset to Courses to create a training course. Testing and exercises validate that your workforce knows exactly how the data should be labeled.
Totally Free, No Sign Up
Get started in less than a minute. Courses uses administrator links. No sign up needed.
View Certified Labelers
View all labelers that have passed, or programmatically query and integrate certification into your workforce system.
Create Tests and Exercises
Create testing where labelers must match the Intersection over Union error of a segmentation, classifications and more.
Jupyter
Label data directly in Jupyter Notebook
Import the Universal Data Tool pip module to label data, create collaborative sessions and transform datasets.