Project 2. Highlight each object by a bounding box

  1. Project creation
  2. Pool creation
  3. Preparing and uploading the file with tasks

Before you start using Toloka, make sure that the English language is selected.

Project creation

  1. Click the + Create project button and choose the Selecting a region in an image template.
  2. Enter a short project name and description, and write simple and clear instructions.
  3. Define parameters for the input and output data:
    • The image input data field with the link type will be used to pass the image links to the performers.
    • The result field with the "json" type is used to record the coordinates of the points marked by the performer.
  4. Create the task interface.
    The image area selection editor is automatically included in the template. This means that the performer will see the image and can select an area with the editor tools.

    You don't have to change anything in the interface. Leave HTML and JS box unchanged.

  5. By default, rectangle tool is hidden in the task interface. To show it, remove the existing code from the CSS field. To hide polygon tool, add the following code:
    .image-annotation-editor__shape-polygon { display: none;}
  6. Add the following code to set the height of the image markup interface to match the image size:
    .image-annotation-editor__annotation-layer { height: max-content;}
  7. Click the Preview button to see the performer's view of the task. Make sure it can be submitted.

    You will see four tasks with standard pictures on the page. You will set the number of tasks per page when configuring a pool.

  8. Click Save button to save the project.

Pool creation

  1. Click Add pool.
  2. Give the pool any convenient name and description. You are the only one who can see them.
  3. Specify the pool parameters:
    • Price per task page (for example, $0.01).
    • Overlap. This is the number of users who will complete the same task. Because everyone will be outlining your traffic lights in their own unique way we will show each photo to only one performer.
    • Time allowed for completing a task page (for example, 600 seconds).
    • Turn on the Non-automatic acceptance option and enter the number of days for checking in the Deadline field (for example, 7).
  4. Set up user filters.
    • Language = EnglishSelect English-speaking performers using the filter.
    • Create the worked_on_bounding skill that will be assigned to users after they complete the pool tasks. You will use this skill to prevent these users from checking tasks in the next project. Click Create skill:
    • Enter the skill name and add a description if desired. You are the only one who will see it. Leave the skill private, as it is by default and click Add.

      Optionally, specify the percentage of top-rated performers in the Speed/Quality ratio.


      This can slow down pool completion.

  5. Set up quality control:
    • Submitted responses. Add a rule to mark users who completed at least one task in the pool.

      If the skill you created doesn't appear in the drop-down list, save the pool, and then open it for editing again.

    • Optionally, add other quality control rules.

      Control tasks and majority vote are not used in this type of project, because user highlighting must exactly match the reference, which is practically impossible.

  6. Save the pool.

Preparing and uploading the file with tasks

  1. Open the file with aggregated results from the Project 1. Does a photo contain an object?.
  2. Select only images suitable for highlighting (OK answers or another value if you have changed it in the result field). Use a text editor or a spreadsheet editor.
  3. Copy the column with the selected links to a new page or document and give a name to the INPUT:image column.

    If you changed the input field name in the project, change the name in the file as well: INPUT:<your field name>.

  4. Save the file in TSV format.
  5. Upload the file to the pool by selecting Set manually. Set 1 task per page.
  6. Start the pool.