Gallery image effects on facial recognition systems

In this paper we quantify the effects of using different types and resolutions of gallery images on two different 2D facial recognition systems. Similar tests have been conducted in this area and adequately reported in the literature, however the galleries used for laboratory testing are usually standardised data sets, the results of which can be misleading if extrapolating into the operational environment. In contrast, our approach uses a data set obtained from operational trials. By testing in such a way, we produce data that can provide more accurate performance estimates across different gallery images for operational facial recognition systems.