From Face Recognition to Models of Identity: A Bayesian Approach to Learning About Unknown Identities from Unsupervised Data

Abstract

Current face recognition systems robustly recognize identities across a wide variety of imaging conditions. In these systems recognition is performed via classification into known identities obtained from supervised identity annotations. There are two problems with this current paradigm: (1) current systems are unable to benefit from unlabelled data which may be available in large quantities; and (2) current systems equate successful recognition with labelling a given input image. Humans, on the other hand, regularly perform identification of individuals completely unsupervised, recognising the identity of someone they have seen before even without being able to name that individual. How can we go beyond the current classification paradigm towards a more human understanding of identities? We propose an integrated Bayesian model that coherently reasons about the observed images, identities, partial knowledge about names, and the situational context of each observation. While our model achieves good recognition performance against known identities, it can also discover new identities from unsupervised data and learns to associate identities with different contexts depending on which identities tend to be observed together. In addition, the proposed semi-supervised component is able to handle not only acquaintances, whose names are known, but also unlabelled familiar faces and complete strangers in a unified framework.Work done during an internship at Microsoft Research.

A Comprehensive Survey on Wi-Fi Sensing for Human Identity Recognition

“…In most CV-based methods, the typical biometric image sequence of a human body is first collected through a camera device; then computer vision computing is adopted to extract and identify human feature information, such as face [43][44][45], fingerprint [46,47], iris [48,49], gait [50,51], etc. In 2019, Wei et al [52] utilized deep learning techniques, employing spatial and temporal attention pooling networks to eliminate redundant information from videos and automatically determine a person's identity.…”

Different Methods of Authentication for Mobile Banking

“…Ref. 16 puts outward an assorted Bayesian model that logically deliberate visualized visuals, identities, incomplete information of names, and the specific context of each remark. The model take care of detect new identities from alone dossier and learn to link identities accompanying various positions contingent upon which identities likely expected viewed together late it achieves good recognition depiction against settled identities.…”

Social Sensemaking with AI: Designing an Open-ended AI Experience with a Blind Child

“…Images are first processed by a model to detect the pose and location of all people in a scene (re-implementation of [12]). Poses are then used to crop the face, which is passed on to a probabilistic model that provides a set of potential identities [14]. The poses are also fed into models that provide information on gaze direction (implementation of [42]) and activity (unpublished).…”