Sure.. there’s the cute iPhone apps like NameTag that make promises of future software enhancements that will make facial recognition software more precise, but have you heard of the GaussianFace algorithm?
Amazingly, this new algorithm is the first ever to identify human faces more accurately than us humans can – with a stunning 98.52% accuracy. Us living, breathing homosapiens came in at just over 97% accuracy.
Not too shabby. Go us!
Facial recognition software has long used a dataset known as LFW, or Labeled Faces in the Wild, to determine accuracy. This dataset contains over 13,000 images of faces – many different mugs and many, the same but with with different expressions, angles and lighting.
The GaussianFace algorithm has achieved accuracy that no other algorithm has even come close to.
You can read more on the details and results of the GaussianFace study here.
Researchers were excited about the results of this new algorithm, pointing to the practical applications of improved security and even using biometric credentials to gain access to computer systems. The government is no doubt excited about the practical applications of greater control over society.
Next up – a new surgery is 97% successful in giving you a new face. Algorithm foiled.