Researchers in Hong Kong have created a facial recognition system that is better than the human eye. Called GuassianFace, the creators hope to one day help police better identify criminals, for gadgets to only unlock when they ‘see’ their owner, or for the system to be used at passport controls, where it would compare your face with the one on your passport.
Check out also: Vegetable Growth Chamber – NASA
Chaochao Lu and Xiaoou Tang, both Chinese University Researchers tested the system using a database known as Labelled Faces in the Wild, which is a collection of 13,000 photos of 6,000 public figures used to test the accuracy of facial recognition systems. GuassianFace was shown pairs of peoples faces and required to decide whether the two people were the same person. In trials, GuassianFace scored 98.5% accuracy, which is better than humans who can perform with up to 97.5% accuracy.
The system is not yet perfect, but researchers have stated that ‘“it still takes a long time to train our model for the high performance. In addition, large memory is also necessary. Therefore, for specific application, one needs to balance the three dimensions: memory, running time, and performance”. By analysing key features of each picture such as the position of the eyes, nose and corners of the mouth, GuassianFace creates a digital map of the face and then overlays it with the other image, looking for similarities, even with very poor quality photos.