State of the Art: Face Recognition

TL;DR

This paper reviews face recognition methods for images with natural and eye occlusions, aiming to identify the most effective baseline approach for recognizing occluded faces, especially in forensic CSEM applications.

Contribution

It provides a comparative review of face recognition techniques tailored for occluded faces, focusing on selecting the best baseline for forensic use cases.

Findings

Identifies challenges in recognizing occluded faces in CSEM.

Evaluates existing face recognition methods for occlusion robustness.

Suggests the most promising baseline approach for occluded face recognition.

Abstract

Working with Child Sexual Exploitation Material (CSEM) in forensic applications might be benefited from the progress in automatic face recognition. However, discriminative parts of a face in CSEM, i.e., mostly the eyes, could be often occluded to difficult the victim's identification. Most of the face recognition approaches cannot deal with such kind of occlusions, resulting in inaccurate face recognition results. This document presents a short review face recognition methods for images with natural and eye occlude faces. The purpose is to select the best baseline approach for solving automatic face recognition of occluded faces.