Studio FAQ

Platform Overview

How do I use Studio? (Platform Overview)

How do I use Quality Tasks?

Are there any data minimums to get started?

No! Studio is completely self-serve and you can label as much, or as little data as you would like.

What data types are supported with Studio?

Studio supports general image annotation, 2D semantic segmentation, text collection, document transcription, named entity recognition, video playback annotation, and lidar annotation.

My data type isn’t supported on Studio but I’d still like to use the platform. What should I do?

Text collection is our most general task type and may work for your use case through the use of presentation templates. Presentation templates allow you to build any task from scratch in markdown via CSV upload where each row corresponds to a task and each column corresponds to some type of attachment (text, image, link, etc) that will be displayed within the task.If your task is more complex and cannot be represented with text collection, please contact our sales team.

What’s the difference between a Calibration Batch, a Regular Batch, and a Self-Label Batch?

  • Calibration Batch: The intention of a calibration batch is to get feedback from your labelers on the instructions you have left for them before you scale up labeling volumes. This is an opportunity for labelers to tell you where your instructions might be unclear - or if there is additional information you need to specify to help them produce better labels. You can submit as many calibration batches until you feel your instructions are in a comprehensive spot. Please note: you do not have to submit a calibration batch if you feel your labelers are already well familiarized on how to label your data.
  • Regular Batch: A regular batch is what you would typically create once you feel confident in your project set-up and you are ready to get your data labeled!
  • Self-Label Batch: You (and other administrators) are the only ones who can label a self-label batch. These are generally used if you want to trial out the labeling platform and before bringing on your team of labelers. Self label batches are not subject to review layers or benchmarks as the assumption is that you - the administrator - are already an expert on creating the best labels. The other use of self label batches is to upload a set of tasks that you want to serve as training tasks or evaluation tasks (benchmarks) for your labelers. Training and evaluation tasks can only be created through auditing - so this is a way for you to create those quality tasks even before you launch a regular batch.





General Questions

How does pricing work? Can I just pay-as-I-go?

After you use up your free trial of 10,000 labeling units, you must commit to purchasing one of 5 different bundles. There is no pay-as-you-go option where you get charged for however many labeling units you use up. If you are running low on labeling units, you can top-up with another bundle purchase.

What is a labeling unit?

Every time one of your labelers annotates a label of any kind, that is considered a labeling unit. A labeling unit is defined as one annotation, or two attributes. An attribute is a descriptive quality that you would want to know about an annotation. For example, if you drew a bounding box around a car and you wanted to know the colour of the car, the bounding box would be the annotation, but the colour would be an attribute.

Do I get charged based on the number of labelers I invite to my project?

No! Feel free to invite as many labeling members as you’d like onto your platform. You only get charged for the work that your labelers complete.

Is there a subscription fee?

No! We only charge based on the number of annotations you would like to buy. No additional charges applied.

How is self-labeling different? Do I need to use Studio, or can my labelers just use self-label?

We recommend you use self-label only if you, or a close team of collaborators plan on doing the labeling. With self-label, you can not permission your members differently and everybody on your project will have access to admin controls. There is also no summary view of performance by labeler. If you are trying to manage any sort of a labeling workforce, we would recommend using Studio.

Updated 3 months ago