The system isn't perfect but promises a success rate of over 90%:
Fraunhofer’s Google Glass app is based on its tried-and-tested SHORE (Sophisticated High-speed Object Recognition Engine) system. SHORE started off as an object-detection computer vision system, but over the years it has developed into a face detection and fine facial analysis system. It can pick out a person’s face with a 91.5% success rate, and tell you that person’s gender 94.3% of the time. It can even take a stab at the person’s age. Previously, SHORE — which is essentially a highly optimized C++ library — has been deployed on various computer systems, from PCs to tablets. Now, Fraunhofer IIS has squeezed all of that facial analysis goodness onto Google Glass’s rather wimpy hardware (1GB of RAM, dual-core TI OMAP 4430 SoC). Watch the video below; it’s a little bit scary how accurate the system is at detecting gender, age, and emotional state.To limit privacy concerns, Fraunhofer chose to never let the data leave the device. The app also can't determine the identity of people that you look at but technically there's nothing preventing someone from creating a similar app that does both of these things.
Source: ExtremeTech