In the case of facial recognition, some of these concerns are due to a lack of awareness of the technology itself.
Facial recognition is increasingly being seen as an excellent tool in improving security. But many customers are still reluctant to adopt this technology. Some of the concerns they share have persisted from its early days while some are relatively new issues that stem from studies conducted on the applications. Addressing them is essential for market growth to continue.
Asmag.com recently talked to some industry professionals to understand the common concerns they encounter from customers and how this should be tackled.
-
Concerns about demographic bias
Recent research has thrown light on certain racial and gender bias concerns that facial recognition algorithms may have. Researchers from MIT and Stanford University who analyzed three commercially released facial-analysis programs showed that error rates in determining the gender of light-skinned men were never worse than 0.8 percent.
However, for darker-skinned women, the error rates ballooned — to more than 20 percent in one case and more than 34 percent in the other two. Several other studies also showed similar concerns that have prompted customers to question the use of facial recognition.
“Primarily, they have heard or read about potential biases in the data upon which the facial recognition algorithms have been trained,” said Kahren Arzoumanian, Co-founder and CTO of Kampr Systems. “This claim applies, if at all, to surveillance solutions where it is an N:1 (many to 1) comparison without much control over the distance and user’s facial lighting and angle. Additionally, rhetoric associated with social and political concerns has forced many integrators to put a hold on any facial recognition technology without knowing the future legal outcomes.”
-
Concerns about its underlying ethics
Ethical concerns about facial recognition technology have plagued the industry since the beginning. This essentially stems from the fact that facial recognition and identification are an invasion of privacy. There are also chances that, in the wrong hands, the use of this technology can have disastrous results.
“Facial recognition has been in the news lately and usually for all the wrong reasons,” explains Dean Nicolls, Chief Marketing Officer of Oosto. “Most of the news stories focus on the use of facial recognition by law enforcement in public surveillance systems. Integrators need to educate their customers about how facial recognition solutions can still protect the digital identities and biometrics of people who come into the video frame. This is typically not a concern with access control use cases.”
-
Concerns about accuracy
Accuracy concerns are another issue that detractors of facial recognition continue to bring up. But most industry professionals are quick to point out that these concerns were mainly an issue in the early days of the technology. Algorithms are now smart enough even to identify faces that are wearing the mask for COVID-19 protection.
“Several years ago, facial recognition was not terribly accurate, especially in real-world scenarios, and the industry is still haunted by these misperceptions,” Nicolls said. “The fact is modern facial recognition is more than 99 percent accurate and performs well even in challenging environments where the lighting is not great, or the person is not looking directly at the camera. This is thanks to the power of today’s AI-based neural networks. A simple demo and proof of concept will quickly alleviate these concerns.”
Also read: Why facial recognition is essential to health care now
-
High perceived costs
Cost is a concern for the use of any technology, especially in the commercial sector. Many customers have voiced concerns that facial recognition may be too expensive for them compared to the current technologies already in place. Again, this is a misconception, according to industry players who point out that costs have come down drastically in recent years as technologies like edge-computing have become mainstream.
“Historically, the cost of facial recognition was high because of the total investment required,” Nicolls said. “But thanks to edge and near-edge compute devices, the TCO is coming down and in pretty dramatic ways (a near-edge compute device can cost as little as one-third per video channel compared to traditional GPU-heavy on-premise servers required to process video footage.”
Conclusion
Addressing customers’ concerns is necessary for any solution provider. In the case of facial recognition, some of these concerns are due to a lack of awareness of the technology itself. Issues such as biases, that researchers point out, should be rectified to ensure user trust.