Abstract
In this work, we examine if current state-of-the-art deep- learning enhanced face recognition systems exhibit a negative bias for children as compared to adults. The systems selected for this work are five top performing1 commercial off-the-shelf (COTS) face recognition systems, two government off-the-shelf (GOTS) face recognition systems and one open-source face recognition solution. The datasets used to evaluate the performance of the systems are both unconstrained in age, pose, illumination, and expression and are publicly available. These datasets are indicative of photo journalistic face datasets published and evaluated on over the last few years. Our findings show a negative bias (i.e., a considerable degradation in performance) for each algorithm on children when compared to the performance obtained on adults. Genuine and imposter distributions high- light the performance bias between the datasets further sup- porting the need for a deeper investigation into algorithm bias as a function of age cohorts. To combat the performance decline on the child demographic, several score- level fusion strategies were evaluated. This work identifies the best score-level fusion technique for child demographic.
