Generating Saliency Maps Using Human Based Computation Agents

摘要

This paper presents an agent system, which uses human based computation (HBC), that is able to calculate a saliency map for a specific image. This system will learn from human interaction to obtain a saliency map of the most important parts of an image. As we will see later, the maps generated using HBC are more robust than their classical counterpart, because they are less dependent on the group of features that exist in a given image.