Object recognition and area selection (example with decomposition)

This is an example of decomposing the Object recognition and area selection task. Decomposition can improve the quality of results and reduce the cost of performing complex tasks.

Let's say you have photos of streets and you need to select traffic signs in them. But you don't know if there are traffic signs in all the photos, so you want to filter them first. You also don't want to check the performers' responses yourself afterwards.

In this example, the solution consists of the following steps:

  • Project 1: Use this project if some images don't contain the intended object and you want to filter them out.
  • Project 2 — in this project, performers will select areas with the desired object in the images. Toloka provides an editor for selecting an image area. It lets the performer select a polygon or rectangle area.
  • Project 3 — this project allows you to ask Toloka performers to review the tasks instead of reviewing them yourself.

Each project consists of the following basic steps:
  1. Create a project. In the project, you describe the input and output data, task interface, and instructions for completing a task.

  2. Create a task pool in the project. In the pool, you set up quality control and filters for performers.

  3. Upload a TSV file with tasks to the pool.

  4. Start the pool.

  5. Obtain and aggregate the results.

Find out how to top up your account in the relevant section: For non-residents of Russia and For residents of Russia.

If you need help in rating tasks, read about setting up pricing and see examples of cost for different types of tasks.