Out of the myriad classes of objects in the world, faces hold special interest for humans. Faces provide a wealth of information about others: identify, age, relatedness, mood, social relationship, etc. However, in spite of substantial progress in computer face recognition over the past two decades, face recognition in natural, uncontrolled environments, remains a largely unsolved problem.
Since biological visual systems are currently unrivaled by artificial systems in real-world face recognition, one natural approach is to build systems that draw inspiration from biology. To this end, we have begun applying our large-scale biologically-inspired feature search approach to the problem of unconstrained face recognition. Our systems currently achieve state-of-the-art performance on the difficult “Labeled Faces in the Wild” face recognition challenge set.
In collaboration with Zak Stone and Todd Zickler (Harvard SEAS), we are taking our exploration of biologically-inspired automated face recognition further, using large-scale, real-life datasets drawn from the Facebook social networking site. Enormous data sets such as these leverage the new scale of data made available by the internet age, and provide powerful new tools for evaluating and advancing our biologically-inspired systems.