The absence of a previous collaborative manual enrolment represents a significanthandicap towards designing a face verification system for face re-identification pur-poses. In this scenario, the system must learn the target identity incrementally, usingdata from the video stream during the operational authentication phase. So, manuallabelling cannot be assumed apart from the first few frames. On the other hand, eventhe most advanced methods trained on large-scale and unconstrained datasets sufferperformance 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 unsupervisedlearning. Our Dynamic Ensemble of SVM is capable of incorporating non-labelled in-formation to improve the performance of any model, even when its initial performanceis modest. The proposal uses the self-training approach and is compared against otherclassification techniques within this same approach. Results show promising behaviourin terms of both knowledge acquisition and impostor robustness.
Keywords: Adaptive biometrics, Video surveillance, Video-to-Video face verification,Unsupervised learning, Incremental learning