Performance variation of morphed face image detection algorithms across different datasets

摘要: In past years, different researchers have shown the vulnerability of face recognition systems to attacks based on morphed images. More recently, first morph detection subsystems been proposed automatically detect this kind fraud. While some algorithms reported reveal practical performance individual datasets a systematic analysis detectors with respect their robustness across databases has remained elusive. work, we evaluate disjoint 2,745 bona fide and 14,337 generated Within generic evaluation framework estimation scheme is identify reliable algorithms. Finally, which determined as most promising verified another dataset. Hence, paper represents attempt towards comprehensive cross-database image