GazeFollow is a large dataset that annotates where people’s eyes are looking in images. Its images come from several datasets containing images of people:

1. From SUN Dataset1,548 images of
2. From MS COCO dataset33,790 images of
3. From Actions 40 Dataset9,135 images of
4. From PASCAL Dataset7,791 images of
5. From ImageNet dataset508 images of
6. From Places dataset198,097 images.

MIT published the dataset in 2015 and it can be used to study what people are looking at in different everyday situations.