Internet Scale Facial Recognition

November 10, 2014

Whatever the privacy qualms, facial recognition software is here to stay and only getting better. (Or worse, depending on one’s perspective.) We’ve found a resource that provides a useful review of algorithms and accuracy: “Computer Vision and Image Understanding” (pdf) is an Elsevier-published paper by the University of Central Florida’s Enrique G. Ortiz and Carnegie Mellon’s Brian C. Becker. Not surprisingly, Facebook photos played a part in the team’s research. The paper’s Abstract explains:

“With millions of users and billions of photos, web-scale face recognition is a challenging task that demands speed, accuracy, and scalability. Most current approaches do not address and do not scale well to Internet-sized scenarios such as tagging friends or finding celebrities. Focusing on web-scale face identification, we gather an 800,000 face dataset from the Facebook social network that models real-world situations where specific faces must be recognized and unknown identities rejected. We propose a novel Linearly Approximated Sparse Representation-based Classification (LASRC) algorithm that uses linear regression to perform sample selection for 1-minimization, thus harnessing the speed of least-squares and the robustness of sparse solutions such as SRC. Our efficient LASRC algorithm achieves comparable performance to SRC with a 100-250 times speedup and exhibits similar recall to SVMs with much faster training. Extensive tests demonstrate our proposed approach is competitive on pair-matching verification tasks and outperforms current state-of-the-art algorithms on open-universe identification in uncontrolled, web-scale scenarios.”

I’m no software engineer so, to be honest, I only understand about two-thirds of the preceding paragraph. However, I’m advised by someone who does know a smart vector from a hole in the ground that this is a resource folks interested in the field should check out. He also points us here for supplemental information.

Cynthia Murrell, November 10, 2014

Sponsored by ArnoldIT.com, developer of Augmentext

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta