On the effect of selfie beautification filters on face detection and recognition

Beautification and augmented reality filters are very popular in applications that use selfie images. However, they can distort or modify biometric features, severely affecting the ability to recognise individuals’ identity even detect face. Accordingly, we address effect of such on accuracy automated face detection recognition. The social media image studied contrast, illumination, occlude parts We observe some these is harmful recognition, especially if obfuscate eye (to a lesser extent) nose. To counteract effect, develop method reverse applied manipulation with modified version U-NET segmentation network. This observed contribute better recognition accuracy. From perspective, employ distance measures trained machine learning algorithms features extracted using several CNN backbones. also evaluate incorporating filtered images into training set approaches beneficial. Our results show good when do not important landmarks, eyes. combined proposed allows mitigating impact produced by