Annotation Suite

Whether you’re solving a single problem or building a company around computer vision, you need Lodestar.

We’ve created a suite of intuitive tools help you unlock the value of your company’s visual data. If you need to turn video into computer vision applications, on our platform you can do it faster and for less money, so you can focus on solving real problems. Lodestar Annotation Suite includes everything you need to become more productive creating high-quality datasets for computer vision projects.

Project Explorer

Monitor Your Project Portfolio

  • Create new projects and get labelers starting by simply sharing a link.
  • See all of your current projects, their number of frames, and number of annotations.
  • See the cross-validation accuracy of each project’s AI model.

Project Dashboard

With everything in one place, you can run more projects and experiments than ever before. Create new projects, manage GPU acceleration, load more video, and monitor labeling and model training progress. Export your labeled datasets, monitor and manage annotation quality, and more.

Video Annotator

Intuitive, no-code interface

  • Easily navigate through hours of video with familiar player controls.
  • Zoom in and out of the frame to verify objects and refine bounding boxes.
  • Create object categories and monitor annotation counts without leaving the annotator.
  • Add notes to annotations.

Zoom Control

Creating high-quality datasets is about precision labels. Quickly zoom in and out of the video frame to find small objects, refine your bounding boxes, and check your work.

Predictive labeling and active learning accelerate projects

  • As you label, Lodestar is training a computer vision model in the background.
  • With only twenty labeled objects it starts predicting labels, helping your labelers work at least three times faster.
  • Active learning powered by a continuously trained AI model suggests the best frames to work on in a gallery below the working frame. Choose one to edit and confirm the machine annotations.
  • Machine annotations display the confidence level the model has in its prediction.

Confidence Level Filtering

  • The Confidence Graph shows at a glance the objects detected in the current and surrounding frames.
  • With a keystroke, filter the machine annotations by confidence level to cut through the noise when the model is still learning.