Towards a self-sufficient face verification system

• Published in: Expert systems with applications 2021-07, Vol.174, p.114734, Article 114734
• Main Authors: Lopez-Lopez, Eric, Regueiro, Carlos V., Pardo, Xosé M., Franco, Annalisa, Lumini, Alessandra
• Format: Article
• Language: English
• Subjects: Adaptive biometrics; Face recognition; Incremental learning; Knowledge acquisition; Performance degradation; Performance enhancement; Target recognition; Unsupervised learning; Verification; Video data; Video surveillance; Video-to-video face verification
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2021.114734

•Self-updating approach for video-to-video face verification tasks.•Proposal of Dynamic ensemble of SVMs as an adaptive biometric system.•Ensembles self-update to target appearance changes due to time or ambient factors.•Incremental learning takes place online when new positive samples emerge.•Results show the viability of the proposal against other analogous approaches. The absence of a previous collaborative manual enrolment represents a significant handicap towards designing a face verification system for face re-identification purposes. In this scenario, the system must learn the target identity incrementally, using data from the video stream during the operational authentication phase. So, manual labelling cannot be assumed apart from the first few frames. On the other hand, even the most advanced methods trained on large-scale and unconstrained datasets suffer performance degradation when no adaptation to specific contexts is performed. This work proposes an adaptive face verification system, for the continuous re-identification of target identity, within the framework of incremental unsupervised learning. Our Dynamic Ensemble of SVM is capable of incorporating non-labelled information to improve the performance of any model, even when its initial performance is modest. The proposal uses the self-training approach and is compared against other classification techniques within this same approach. Results show promising behaviour in terms of both knowledge acquisition and impostor robustness.