This FiFa database is dedicated to sharing data pertaining to observers viewing of faces in natural scenes. The data provided includes subjects’ viewing of images containing faces, showing that faces are attracting the attention. Fixations are provided in Matlab format. Additionally, a Matlab code used for determining the saliency of locations in given images is provided. This code is used to predict the locations of subjects fixations in images.
Below is a link to the fixations of 8 subjects viewing images of natural scenes. Link: http://www.klab.caltech.edu/~moran/fifadb/fixations.mat In addition to the fixations data, we provide an annotation of the entire dataset. That is, the location and labeling of faces in images are given. This allows for easier further analysis of the data. Link: http://www.klab.caltech.edu/~moran/fifadb/annotations.mat All the images used in the study can be viewed. Link: http://www.klab.caltech.edu/~moran/db/faces/faces-tif.tgz For a list of all images used in the study (corresponding to the fixations struct) download the list of images below. Link: http://www.klab.caltech.edu/~moran/fifadb/imgList.mat
References and Citation
Please acknowledge usage of this data by citing this paper [CFK09]. Additionally, please refer to the paper for details on the fixations and the method by which they were acquired. Please acknowledge usage of these images by citing this paper [CHE07]. Additionally, please refer to the paper for details on the images and the method by which they were acquired and used in prior studies.
CFK09: Moran Cerf, Paxon Frady, Christof Koch, Faces and text attract gaze independent of the task: Experimental data and computer model, Journal of Vision, 9(12):10, 1-15, 2009.
CHE07: Moran Cerf, Jonathan Harel, Wolfgang Einhaeuser, Christof Koch, Predicting human gaze using low-level saliency combined with face detection, Advances in Neural Information Processing Systems (NIPS), 2007.