Is the eye region more reliable than the face? A preliminary study of face-based recognition on a transgender dataset

In this work we investigate a truly novel and extremely unique biometric problem: face-based recognition for transgender persons. A transgender person is someone who under goes a gender transformation via hormone replacement therapy; that is, a male becomes a female by suppressing natural testosterone production and exogenously increasing estrogen. Transgender hormone replacement therapy causes physical changes in the body and face. This work provides a preliminary investigation into the effects of these changes on face recognition systems: commercial matcher as well as established texture-based matchers (LBP, HOG, SIFT). The performance of the full-face matchers are compared with that of the periocular region for the same feature sets. The results indicate that periocular recognition out performs the full face on the transgender dataset under real world conditions. In addition, we introduce a novel dataset for researchers: transgender dataset, which was organized from public sources.