Rapid progress in unconstrained face recognition has resulted in a saturation in recognition accuracy for current benchmark datasets. While important for early progress, a chief limitation in most benchmark datasets is the use of a commodity face detector to select face imagery. The implication of this strategy is restricted variations in face pose and other confounding factors. This paper introduces the IARPA Janus Benchmark A (IJB-A), a publicly available media in the wild dataset containing 500 subjects with manually localized face images. Key features of the IJB-A dataset are: (i) full pose variation, (ii) joint use for face recognition and face detection benchmarking, (iii) a mix of images and videos, (iv) wider geographic variation of subjects, (v) protocols supporting both open-set identification (1:N search) and verification (1:1 comparison), (vi) an optional protocol that allows modeling of gallery subjects, and (vii) ground truth eye and nose locations. The dataset has been developed using 1,501,267 million crowd sourced annotations. Baseline accuracies for both face detection and face recognition from commercial and open source algorithms demonstrate the challenge offered by this new unconstrained benchmark.

This paper studies the influence of demographics on the performance of face recognition algorithms. The recognition accuracies of six different face recognition algorithms (three commercial, two nontrainable, and one trainable) are computed on a large scale gallery that is partitioned so that each partition consists entirely of specific demographic cohorts. Eight total cohorts are isolated based on gender (male and female), race/ethnicity (Black, White, and Hispanic), and age group (18–30, 30–50, and 50–70 years old). Experimental results demonstrate that both commercial and the nontrainable algorithms consistently have lower matching accuracies on the same cohorts (females, Blacks, and age group 18–30) than the remaining cohorts within their demographic. Additional experiments investigate the impact of the demographic distribution in the training set on the performance of a trainable face recognition algorithm. We show that the matching accuracy for race/ethnicity and age cohorts can be improved by training exclusively on that specific cohort. Operationally, this leads to a scenario, called dynamic face matcher selection, where multiple face recognition algorithms (each trained on different demographic cohorts) are available for a biometric system operator to select based on the demographic information extracted from a probe image. This procedure should lead to improved face recognition accuracy in many intelligence and law enforcement face recognition scenarios. Finally, we show that an alternative to dynamic face matcher selection is to train face recognition algorithms on datasets that are evenly distributed across demographics, as this approach offers consistently high accuracy across all cohorts.

We are actively seeking to increase both the number and competitiveness of proposals to NSF's Course, Curriculum, and Laboratory Improvement (CCLI) program. Funding for computer science education programs at NSF responds directly to the total proposal pressure, so increasing both the number of proposals and the competitiveness of those proposals increases the amount of funding NSF targets for computer science education.

No abstract available.

A class of biometrics based upon ear features is introduced for use in the development of passive identification systems. The viability of the proposed biometric is shown both theoretically in terms of the uniqueness and measurability over time of the ear, and in practice through the implementation of a computer vision based system. Each subject's ear is modeled as an adjacency graph built from the Voronoi diagram of its curve segments. We introduce a novel graph matching based algorithm for authentication which takes into account the erroneous curve segments which can occur due to changes (e.g., lighting, shadowing, and occlusion) in the ear image. This class of biometrics is ideal for passive identification because the features are robust and can be reliably extracted from a distance.

We address the problem of describing, recognizing, and learning generic, free-form objects in real-world scenes. For this purpose, we have developed a hybrid appearance-based approach where objects are encoded as loose collections of parts and relations between neighboring parts. The key features of this approach are: part decomposition based on local structure segmentation derived from multi-scale wavelet filters, flexible and efficient recognition by combining weak structural constraints, and learning and generalization of generic object categories (with possibly large intra-class variability) from real examples.

We address the problem of describing, recognizing, and learning generic, free-form objects in real-world scenes. An appearance-based system using weak-structure and evidence accumulation where object models are implicitly encoded in a learned decision tree and objects are represented in graph form as in the method developed by Bischof and Caelli (1994). The decision tree is used to classify sequences of image components, or part paths, extracted from the object to be recognized. The part paths are in turn used to accumulate evidence for the classification of the entire object. We introduce an improved method for generating part paths based upon the part compatibility graph, a replacement for Bischof and Caelli's implicit use of the part adjacency graph. A new formalism for extending the representation and recognition scheme to utilize multiple (polymorphic) types of primitive parts is presented and the approach is demonstrated on a selection of imagery.