Topic: Impact of Selfie Filters on Face Recognition Systems
by Cristian Botezatu
online Big Blue Button Room: D19/2.03a, July 07, 2021 (Wednesday), 12.00 noon
Keywords — Face Recognition, Beautification, Selfie Filters
Among other biometric systems, face recognition (FR) is widely accepted, convenient and accurate, seeing significant performance enhancement since the appearance of deep learning. Despite the nearly perfect recognition performance of state-of-the-art FR systems, many research works have raised concerns with regards to their reliability when being exposed to occluded faces.
Following now-a-days trends, especially among young people, an increased demand on mobile applications for selfie filtering is registered, making many social media images unrecognizable. Hence, the project focuses on defining the impact of selfie filters on FR performance, extending the work with a proposed model to remove the selfe filter for an incremental growth in FR performance on state-of-the-art FR systems.
Obtained results show that selfie filters affect both the tested commercial and open-source systems, especially when the facial coverage of the selfie filter increases. Moreover, some of the tested face recognition and detection algorithms have shown to be particularly vulnerable to selfie filters altering the eye, mouth or total face regions.