Can Facial Uniqueness be Inferred from Impostor Scores? Abhishek Dutta Oct 14, 2013, Nijmegen, Netherlands. Presentation accompanying the following paper: A. Dutta, R. Veldhuis and L. Spreeuwers. Can Facial Uniqueness be Inferred from Impostor Scores?, BTFS 2013, Nijmegen, Netherlands.
A quick question Image Pair 1 Are they same subject ?
A quick question ... Image Pair 2 What about this image pair? Are they same subject?
Facial Uniqueness (or, Distinctness) in a Population Image Pair 1 Image Pair 2 person-id person-id person-id 261 263 272 Irrespective of your decision { yes, no, maybe } , you were: ◮ more uncertain in your decision for Image Pair 1, ◮ less uncertain in your decision for Image Pair 2. Why ?
Facial Uniqueness (or, Distinctness) in a Population ... Image Pair 1 Image Pair 2 person-id person-id person-id 261 263 272 If we assume that faces form a Gaussian cloud in a high dimensional space, then ◮ identity relates to direction, and ◮ distinctness relates to distance from the mean. Therefore, we are more confident in making decision about identity when facial images are well separated in the face space – by virtue of their distinctness (as in Image pair 2).
Facial Uniqueness (or, Distinctness) in a Population ... ◮ Therefore, in Biometrics, researchers have been actively involved in measuring uniqueness from facial photographs ◮ Such facial uniqueness measurements are useful to build an face recognition system that adaptively vary the decision threshold to improve recognition performance.
Measuring Facial Uniqueness how unique is this facial Impostor Population image? with respect to this impostor population? fairly common highly distinct face face count count similarity score similarity score
Measuring Facial Uniqueness ... how unique is this facial Impostor Population image? with respect to this impostor population? fairly common highly distinct face face count count similarity score similarity score Assumption ◮ similarity score is influenced only by facial identity ◮ non-unique facial image 1 will generally exhibit high level of similarity to many other subjects in a large population (by definition) Hypothesis facial uniqueness of a subject can be inferred from its impostor similarity score distribution 1 lamb in the biometric zoo
Can Facial Uniqueness be Inferred from Impostor Scores? In this paper, we investigate the stability of facial uniqueness measures based on impostor scores. For this, we perform following two experiments: Experiment 1 To study the influence of image quality variations on the impostor score distribution. Experiment 2 To study the stability of a recently proposed Impostor Based Uniqueness Measure (IUM) of [Klare and Jain 2013] under image quality variations.
Exp. 1: Influence of Image Quality on Impostor Scores Baseline Gallery Baseline Impostor Score Distribution Pose (camera-id) 08_1 19_1 Probe (Average Face) 08_0 13_0 14_0 05_1 05_0 04_1 19_0 Gaussian Noise Motion Blur (mean= 0) (angle = 0) Quality based Impostor Score Distribution var. = 0.07 var. = 0.3 len. = 09 len. = 31 Gallery with Image Quality Variation
Recommend
More recommend
