Video annotation suite

Our video annotation tool is the fastest, easiest way to create high-quality datasets.

Built for video

No data prep and curation

No data prep and curation

Drag and drop videos right into your projects. No need to convert frames to images and spend days curating them before you start labeling.

Real-time active learning

Real-time active learning

As you label, we continuously train a custom AI that predicts labels and suggests the most valuable labels to add next. As accuracy accelerates, so does labeling.

Hours of video

Hours of video

Add entire videos into your project and get to work. Let the active learning system built into our video labeling tool find the objects in all that video. You're a data scientist, not a video editor.

Drag. Drop. Label.

An intuitive UI and powerful AI assistance make anyone an expert annotator.

An intuitive UI and powerful AI assistance make anyone an expert annotator.
An intuitive UI and powerful AI assistance make anyone an expert annotator.
  • Label 10 hours of video
  • Over 1 million labels per project
  • All plans include Unlimited labels

Continuously-trained AI assistants

Lodestar continuously trains a custom AI model on the live dataset as labels are added. The model drives real-time active learning and custom predictive labeling — two powerful AI technologies that dramatically accelerate high-quality dataset creation.

Active learning

Most valuable labels first

Most valuable labels first

Higher accuracy with fewer labels. Real-time active learning searches for objects to label that will increase model accuracy — underrepresented categories, partially obscured objects, and other edge cases buried in your data.

Custom predictive labeling

Label 4x faster

Label 4x faster

As you add labels, we continuously train a custom AI model. This AI assistant then automatically labels new objects, helping annotators complete complex frames in seconds. This virtuous circle of label, train, and predict supercharges every project.

Powerful yet easy to use

Lodestar is a complete solution for turning raw video into a trained computer vision model. With its intuitive interface for annotation and one-stop-shop suite of integrated data science tools, domain experts and data scientists alike will be more productive.

  • Localization made easy

    The sweet spot for computer vision dataset creation. When localization is as easy as classification, all of your datasets can be more valuable.

  • Annotation Explorer

    A complete set of tools to QA annotations, document workflows, create exemplars, filter and tag labels, and more.

  • Project Management

    Create new projects, manage GPU acceleration, and monitor labeling and model training progress. Invite new collaborators to projects and manage your workforce.

  • Dataset Management

    Annotators and data scientists can collaborate on a single-source-of-truth dataset in real-time. Random pixel-level access and native video data structures eliminate frame-based workflows and open up new possibilities for your data science experiments.

    Learn more
  • Attribute tagging

    Add object-level attribute tags while labeling. Through our pyLucy API, data scientists can access this metadata in their experiments.

Localization made easyAnnotation ExplorerProject ManagementDataset ManagementAttribute tagging

Frequently asked questions

  • Do you support image labeling?

    Yes. Our video labeling tool supports image datasets as well as video datasets.

  • What video and image formats are supported?

    Currently our platform supports the following formats:
    Video: .mp4, .mov, .avi, .mkv
    Images: jpg, jpeg, png, bmp
    If you have a format of the data that is not in the list, please contact sales.

  • Can I annotate more than 10 hours of video?

    10 hours is not a hard limit for our video annotation platform. Contact us for more information.

  • Unlimited labels? Seriously?

    Create as many labels as you want with any of our plans. See pricing for more details.