10. Job Design

CML (Custom Markup Language)

CML creates the HTML template that will define how your human annotators interact with images from your dataset, and ultimately how they select from different annotations. The parking sign CML is pasted below, for your reference:

<div class="row-fluid">
  <div class="span6">
    <img class="liquid-image" src="{{image_url}}" />
  </div>
  <div class="span6">
    <cml:radios label="Is there a parking sign in this image?" validates="required" name="is_there_a_parking_sign_in_this_image" gold="true">
      <cml:radio label="Yes" value="yes"></cml:radio>
      <cml:radio label="No" value="no"></cml:radio>
    </cml:radios>
  </div>
</div>

You can see that this produces a simple "radio" quiz where annotators will see an image_url, a question about the content of the image and then two radio buttons: yes or no, that will define the image as having a parking sign in it or not.

When creating a custom annotation job, I find it easiest to use a template CML and then modify it for a specific use case.