A new perceptual hashing method for verification and identity classification of occluded faces

作者：

Highlights：

• Victims in Child Sexual Exploitation Material (CSEM) often present the eye region occluded.

• Deep learning face recognition models fail when the eye region is occluded.

• We present a perceptual hashing method called OSF-DNS.

• OSF-DNS is robust against eye occlusion for face verification and classification.

• OSF-DNS outperforms other perceptual hashing and state-of-the-art face recognition methods.

摘要

•Victims in Child Sexual Exploitation Material (CSEM) often present the eye region occluded.•Deep learning face recognition models fail when the eye region is occluded.•We present a perceptual hashing method called OSF-DNS.•OSF-DNS is robust against eye occlusion for face verification and classification.•OSF-DNS outperforms other perceptual hashing and state-of-the-art face recognition methods.

论文关键词：Face verification,Face classification,Adversarial attack eye region occlusion,Perceptual hashing,Discrete cosine transform (DCT),OSF-DNS

论文评审过程：Received 4 March 2021, Revised 20 May 2021, Accepted 13 June 2021, Available online 26 June 2021, Version of Record 6 July 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104245