The Contribution of Facial Components to Face Recognition
Nova Hadi Lestriandoko is a PhD student in the department Datamanagement & Biometrics. (Co)Promotors are prof.dr.ir. R.N.J. Veldhuis and dr.ir. L.J. Spreeuwers from the faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente.
Facial components, such as the eyes, nose, and eyebrows, play a critical role in biometric and forensic identification. While previous studies have explored the effects of partial occlusion and masking on face recognition, the systematic analysis of individual facial components’ contributions to deep face recognition remains limited. This thesis introduces a method to analyze the importance of different face components to face recognition and a tool that enables the selective replacement of facial components to study their impact on recognition performance. Through a series of experiments, we demonstrate that certain components, particularly the eyebrows and nose, carry more discriminative information than others, with texture generally being more influential than shape, except for the eyebrows. The method also extends to historical facial analysis, such as identifying Roman emperors from sculptures by isolating features like hairstyles and facial hair. Moreover, we present an open-source tool developed in Python, capable of seamlessly replacing facial components, including their textures and shapes, suitable for both research and application contexts. This work bridges a gap in facial analysis by providing a systematic approach and practical tool for studying the role of facial components in face recognition.
