Intel Tests Facial Recognition Algorithms on Thermal Images

Intel researchers published a study in which they looked into whether artificial intelligence (AI) algorithms trained to identify individuals’ faces can also be applied to thermal images of faces.

Thermal imaging is seen by many as a way to protect an individual’s privacy, as it obscures some of the details of an image such as eye colour. In medical facilities for example, it’s often required to use thermal images for privacy reasons.

“Many promising visual-processing applications, such as non-contact vital signs estimation and smart home monitoring, can involve private and or sensitive data, such as biometric information about a person’s health,” wrote the researchers, adding that “[t]hermal imaging, which can provide useful data while also concealing individual identities, is therefore used for many applications.”

For their study, Intel’s researchers used two different data sets. The first, known as SC3000-DB, contained 766 images of 40 volunteers, consisting of 21 women and 19 men, each of whom sat in front of an infrared camera for two minutes to capture the images.

The second set, know as IRIS and created using the Visual Computing and Image Processing Lab at Oklahoma State University, contains 4,190 images of 30 individuals. The IRIS dataset differed from the SC3000-DB in that it included various head angles and expressions.

Both datasets were put through a machine learning model which then looked to numerically label facial features from the images as vectors.

The results of the study showed that the model was able to positively identify the volunteers with a 99.5 percent accuracy for SC3000-DB, and 82.14 percent accuracy for the IRIS dataset.

Sources: AI News, Intel