Project Overview
Initially, we prepared an experiment setting with 492 still images from existing datasets. Half of the images were chosen from EMotional Attention Dataset (EMOd) ↗ and the other half of the images were chosen from Saliency in Crowd (EyeCrowd) dataset ↗. Our experiment setting is designed as between-subjects; in total, we have four groups of participants, three of which perform three assigned tasks on the same 492 images, whereas one group is asked to free-view the same 492 images. We also referred to free-viewing as a task (the task of no task). The dataset is publicly available.
- For task 1, we asked the participants to count people in the frame, if any, as they can, and come up with a number, the number does not necessarily have to be very exact, the amount that they could count.
- For task 2, we asked the participants to find an answer for emotion in the frame and people’s feelings in the frame. Before the experiment, some examples are given as 'sad', 'happy', 'angry', etc. and they were also free to choose different words for the emotions as they can think of during viewing.
- For task 3, the participants were asked to define the occasion in the frame, if any, or what person/people in the frame is/are doing.
- Task 4 is free-viewing in which the participants were not asked for any task they were told to view shown images, freely.