Face recognition vendor test 2002 :

Abstract

The Face Recognition Vendor Test (FRVT) 2002 is an independently administered technology evaluation of mature face recognition systems. FRVT 2002 provides performance measures for assessing the capability of face recognition systems to meet requirements for large-scale, real-world applications. Ten commercial firms participated in FRVT 2002. FRVT 2002 computed performance statistics on an extremely large data set-121,589 operational facial images of 37,437 individuals. FRVT 2002 1) characterized identification and watch list performance as a function of database size, 2) estimated the variability in performance for different groups of people, 3) characterized performance as a function of elapsed time between enrolled and new images of a person, and 4) investigated the effect of demographics on performance. FRVT 2002 shows that recognition from indoor images has made substantial progress since FRVT 2000. Demographic results show that males are easier to recognize than females and that older people are easier to recognize than younger people. FRVT 2002 also assessed the impact of three new techniques for improving face recognition: three-dimensional morphable models, normalization of similarity scores, and face recognition from video sequences. Results show that three-dimensional morphable models and normalization increase performance and that face recognition from video sequences offers only a limited increase in performance over still images. A new XML-based evaluation protocol was developed for FRVT 2002. This protocol is flexible and supports evaluations of biometrics in general.

Pose Invariant Face Recognition Under Arbitrary Illumination Based on 3D Face Reconstruction

“…And so far, the accuracy of face recognition for frontal face with indoor lighting is very high [7]. But in some daily applications, the recognition tasks are difficult for the uncontrolled variations in lighting and pose.…”

“…The 3D faces can be generated automatically from one or more photographs by optimizing the shape parameters, the texture parameters and the mapping parameters. This morphable method has been used in FRVT 2002 for its good performance [7]. But the iterative optimal procedure causes the time consuming and the fitting process takes 4.5 minutes on a workstation with a 2Ghz P4 processor.…”

Is Enough Enough? What Is Sufficiency in Biometric Data?

“…We would like to look at some of the them from the point of the view of statistical significants of the results using formula (18). The databases considered are LDB describing gait from the University of Southampton mentioned above [12], CASIA iris image database [4,2], FERET database of facial images [6], subset of BEN database of fingerprint images [14] and FRVT2002 database of faces [7]. CCRs reported for these databases and a number of samples used to obtain these CCRs are presented in Table 2.…”

Bilinear Active Appearance Models

“…In particular, uncontrolled face-pose scenarios correspond to situations that cause current commercial face recognition systems to fail, as in the Facial Recognition Vendor Test 2002, for example [31] [20]. This specially is a major inconvience in those applications using only frontal faces as gallery images and probe images captured in any random pose (e.g.…”

“…To gauge how well our proposed pose normalization scheme improves facial recognition, we compared the recognition performance of commercial system FaceIt with and without pose normalization. We chose to use the FaceIt system because it performed best in the Face Recognition Vendor Test 2002 [31]. The Multi-PIE database has 337 subjects in 15 poses and 18 illuminations.…”