Guide to: Line and Dot Image Annotation


Declaring the CML Image Annotation Element

<cml:image_annotation> </cml:image_annotation>

This CML tag renders an image and corresponding tools for image annotation. There are three forms of this image annotation tool:

  • Dot
  • Line
  • Dot and Line

The Dot and Line tools can be used together within the same job, but it is recommended for simplicity to use just one at a time.

The following section describes each attribute of the tool's supported declarations.


Name the Tool

To define the name of the image annotation tool, use the name attribute.

<cml:image_annotation name="image_annotation_instance"></cml:image_annotation>

Note: The name is required for all instances of the cml:image_annotation tool.

Define the Tool Type

You can specify which annotation tool you wish to use by using the ia_tools attribute:

<cml:image_annotation ia_tools="dot"></cml:image_annotation>

<cml:image_annotation ia_tools="line"></cml:image_annotation>

<cml:image_annotation ia_tools="dot line"></cml:image_annotation>
Define the Image to be Annotated

To define the data column that contains the URL of the image to be annotated, use the data_image_url attribute. This field is required for all instances of the cml:image_annotation tool.

<cml:image_annotation data_image_url="{{image_url}}"></cml:image_annotation> 
Enable Test Questions

To define whether test questions will be used for image annotation use the gold attribute tag -- using test questions is highly recommended:

<cml:image_annotation gold="true"></cml:image_annotation> 
Define Matching Methodology

To define the method by which contributors' responses will be checked against test questions, use the matcher attribute. Dots and lines use the greedy_point matcher. The matcher parameter is required when test questions are used.

Greedy Point Matcher
<cml:image_annotation ia_tools="dot" gold="true" matcher="greedy_point"></cml:image_annotation>

The greedy_point algorithm uses two tolerances for rigidness of matching:

  • The point distance tolerance is the maximum acceptable distance in pixels between Dot/Line pairs.
  • The point difference tolerance is the maximum acceptable difference between test question and judgment Dots/Lines.
Define Aggregation Methodology

To define the aggregation method, use the aggregation attribute tag. Dots and Lines use the greedy_agg method.

<cml:image_annotation ia_tools="line" gold="true" aggregation="greedy_agg"></cml:image_annotation>

An example of the full declaration of an image annotation tool is show below:

<cml:image_annotation name="ia" ia_tools="dot" data_image_url="{{image_url}}" gold="true" matcher="greedy_point" aggregation="greedy_agg">


Was this article helpful?
1 out of 2 found this helpful

Have more questions? Submit a request
Powered by Zendesk