This is a summary of what system does. In the beginning the user makes a selection and draws a sketch, the sketches are used by the interface, the interface has access not only to the image, but also to segmentation. So it looks at the segmentation, it looks at the sketch that the user drew, and combines those two pieces of information to select an object, which is the best match of the sketch with the segmentation and gives that at the selection which the user sees and says if it is right or wrong.

---