A team of researchers from the University of Bradford has discovered that facial recognition technology works even when only half of a face is visible. With the use of artificial intelligence techniques, the researchers achieved 100 percent recognition rates for both three-quarter and half faces. The team published the study in Future Generation Computer Systems, and it is the first to use machine learning to test the recognition rates for different parts of the face.

Professor Hassan Ugail from the University of Bradford, the lead researcher, said that the ability of humans to recognize faces is amazing yet study has revealed that it begins to falter when we can only see parts of a face. Computers can already perform better than humans in recognizing one face from a large number; as such, the researchers wanted to know if they would be better at partial facial recognition as well.

Using a 'convolutional neural network,' a machine learning technique, the team drew on a feature extraction model call VGG, one of the most popular and widely used for facial recognition. Then, the researchers worked with a dataset containing multiple photos, about 2800 in total, of 200 students and staff from FEI University in Brazil, with equal numbers of men and women.

The researchers trained the model using only full facial images for the first experiment. Then, they ran a test to see how well the computer was able to recognize faces, even when shown only part of them. The machine recognized full faces 100 percent of the time, but the team also had 100 percent success with three-quarter faces and with the top and right half of the face. The bottom half of the face, however, was only correctly recognized 60 percent of the time and eyes and nose on their own, just 40 percent.

After the team trained the model using partial facial images, then, they reran the experiment. This time, the scores significantly improved for the bottom half of the face, for eyes and nose on their own and even for faces with no eyes and nose visible, achieving around 90 percent correct identification.

Individual facial parts including the nose, cheek, forehead or mouth had low recognition rates in both experiments.

Professor Hassan explained the result further that they have shown that it is possible to have very accurate facial recognition from images that only show part of a face and they have identified which parts are more useful. This experiment opens up greater possibilities for the use of technology for security or crime prevention.

Hassan concluded that their experiments now require validating on a much larger dataset. However, in the future, it is likely that image dataset used for facial recognition will need to include partial images as well so that the models can be trained correctly to recognize a face even when not all of it is visible.