Linear Feature Learning for Kinship verification in the wild
Abdellah Sellam, Hamid Azzoune · 2018 International Conference on Applied Smart Systems (ICASS) · 2018
Kinship verification based on face images in the wild is a new subject that is gaining more attention by the research community in these last years. A variety of approaches have been proposed, however, the problem is still open for further improvements. In this paper, we present a Linear Feature Learning (LFL-KIN) approach that learns a linear transformation of raw pixel data from the images of the two subjects (parent/child) into a new space where images of positive pairs are more similar than those with no kin relationship. We focus in this study on the number of extracted features and its impact on the performance of the resulting model. The proposed LFL-KIN approach was able to achieve accuracy competitive with state of the art approaches.