Identifying ethnicity problematic so scientists write ‘race recognition’ code

An unusually diverse team of researchers say they have created the better human race recognition system.

There has been some debate about facial recognition’s role in ethnicity and racial profiling. In China, for example.

Most people not deeply involved in biometrics recognition would never know there is an ethical need for AI ethnicity identification, much less a struggle to find the best way to accomplish the task.

In a new open-access research paper, eight scientists from China, Iraq, Australia and Spain say they tinkered with field programable gate arrays (FPGAs) and a new approach to deep convolutional neural networking to push ethnicity-spotting accuracy to 96.9 percent.

What is more, the team was able to significantly reduce the energy needed to achieve their results.

The experiment used 3,141 photographs of people from Pakistan, China and Russia (perhaps coincidentally, three nations with nationalistic conflicts in recent history). Faces were used as the sole factor in identifying ethnicity, according to the paper.

Four pre-trained convolutional neural network models — GoogleNet, AlexNet, DenseNet and ResNet50 were used for comparing the experimental network.

The hardware challenge was between graphical processing units and the gate arrays, both parallel architectures.

Arrays’ ability to customize hardware implementation proved a “core” advantage over either GPUs or CPUs.

Field programable gate arrays ultimately provided “more significant energy saving” and faster processing, according to the report.

Race and ethnicity recognition today present “significant” challenges in human-computer interaction, the authors write. And successfully employing ‘soft biometric’ systems would streamline identification tasks searching large caches of still and video images.

Gene fragments can result in facial features melding across ethnicities, making it hard for physical anthropologists to determine race, something a sharp race identification algorithm could help resolve, they claim.

The code also could find a role in the growing field of race-targeted medical treatments and pharmacogenomics, where accurately ascertaining race could provide better care.

It could be helpful to some employers. Such as system could “use racial information to offer employers ethnically convenient services, then preventing the offending risk present in many cultural taboos.”

Ultimately, however, the broadest potential mission for race recognition would be in security — at border stations and deployed in public-access areas, according to the report.

Article Topics

accuracy  |  AI  |  biometric identification  |  biometrics  |  ethics  |  ethnicity recognition  |  face biometrics  |  research and development  |  surveillance