Michal Balazia

Face Uniqueness

Face recognition has been widely accepted as a means of identification in applications ranging from border control to security in the banking sector. Surprisingly, while widely accepted, we still lack the understanding of uniqueness or distinctiveness of faces as biometric modality. In this work, we study the impact of factors such as image resolution, feature representation, database size, age and gender on uniqueness denoted by the Kullback-Leibler divergence between genuine and impostor distributions. Genuine and impostor score distributions in a setting with relatively unique subjects (left), as well as a setting with similar subjects (right). Towards understanding the impact, we evaluate the datasets AT&T, LFW, IMDb-Face, as well as ND-TWINS, with the feature extraction algorithms VGGFace, VGG16, ResNet50, InceptionV3, MobileNet and DenseNet121, that reveal the quantitative impact of the named factors.

The amount of biometric information is influenced by a set of factors including facial expression, pose, image resolution, distortion, noise or blur. Moreover, the general population includes family members, a large number of twins (3% in the US between 2014–2018), as well as doppelgangers, all of which inherently lower the overall uniqueness of faces in a dataset.

We present preliminary results on the impact of these factors on facial uniqueness. We provide clear experimental evidence of decrease in the uniqueness score, in the case that (a) image resolution decreases, (b) a single gender is observed, (c) a smaller age group is observed, (d) a larger dataset is used, as well as (e) different feature extractors are used. We illustrate that while feature representation and dataset size significantly affect the uniqueness score, image resolution has a negligible impact. Further, we propose an alternative uniqueness estimate, which reflects on the presence of twins. While these are early results, our findings indicate the need for a better understanding of the concept of biometric uniqueness and its implication on face recognition.

The work was published once in a conference proceedings.