Documentation

Projects and Pipelines: Setting Up Your Data for Annotation

This section explains how to prepare your workspace and data for annotation – from creating a project to launching your first annotation pipeline. Once the setup is complete, you can continue to Section 5 – How to Annotate to learn how to create and refine your annotations.

Projects

A Project serves as your primary place of work. It acts as the container where the actual annotation takes place and where you launch your automation pipelines. Because a project holds the specific configurations for your work – such as defined vocabularies and automations – we recommend creating a separate project for each individual use case to keep your workflows organized and efficient.

Create a Project

  1. Navigate to Annotation → All Projects in the left sidebar. (If this is your first time, click Get Started).
  2. Click the New Project button in the upper right corner.
  3. Enter a Name and Description for your project.
  4. Click Create.

Configure Vocabulary

Recomended

While you can start a project without a pre-defined vocabulary, we highly recommend setting one up if your use case requires specific object categories or activity steps.

Why should you Link a Vocabulary?

  • Speed: It accelerates manual work by providing your defined categories as "suggestions" while you annotate.
  • Automation: It unlocks advanced automations, such as automated activity creation, which rely on a structured vocabulary to function correctly.

Add Videos to Your Project

Once your project is set up, you need to import the video data you wish to annotate.

  1. Navigate to Library → Videos.
  2. Select the videos you want to annotate by checking the box in the top right corner of each video card.
  3. Click the Green Button located in the top right corner of the screen.
  4. In the menu that appears on the right, click Add to Project.
  5. Select your desired project from the list.
  6. Add to project.

Pipelines

Pipelines are the engine of the platform. They use Machine Learning (ML) models to automate large parts of the annotation process within your project, minimizing manual input. We offer pre-built Pipeline Templates to help you get started quickly.

How to Start a Pipeline

  1. Navigate to the Annotation section and open the Project containing your videos.
  2. Click on the Videos tab.
  3. Locate the video card you wish to process and click New. A pop-up window will appear.
  4. In the Select a Pipeline field, choose the template that matches your content:
    • Hand-Interaction Video Annotation: Choose this for videos containing hands interacting with objects.
    • Video Annotation: Choose this for all other video types.
  5. Click Start.

Pipeline Status

  • In Progress: Immediately after starting, the pipeline status will read "In Progress." This means the automation is running in the background.
  • Start Task: Once the automation is complete, a green Start Task button will appear on the annotation card. Click this to open the video and begin your manual review or refinement.

Automations and Customization

Automating Pipeline Creation

You can configure your project to trigger pipelines automatically. This saves time by removing the need to manually click Start for every new video added to the work station.

  1. Open your Project and navigate to the Automations tab.
  2. Click Create Automation.
  3. Select the Pipeline Template you wish to apply.
  4. Confirm the trigger.

Note: Currently, the automation triggers automatically whenever a new document/video is added to the project.

Custom Pipeline Templates

Our pipelines are modular, meaning different machine learning components can be mixed and matched to suit complex workflows. If the pre-built templates do not meet your specific needs, we can help you design a custom solution. Please reach out to our team to discuss creating or modifying a custom pipeline template.