Facial memorability prediction fusing geometric and texture features

As different faces have different features, the degree of memorability of faces are different, which are named memorability in this paper. We mainly study the relation between the memorability and different features such as the geometrical features of the faces, the location of eyes, the size of mouth and eyes and the Histogram of Oriented Gradient (HOG). We use SVR model to regress the features of face images, and predict the memorability score. Finally, we use the spearman rank correlation coefficient and residual sum-of-squares error to analyze the correlation and error of the predicted memorability score with ground truth.