Short answers, pulled from the story.
A facial recognition system matches a human face from a digital image or video frame against a database of faces using four steps: face detection, image alignment, facial feature extraction, and database matching. The technology is categorized as biometrics because it measures a person's physiological characteristics.
Automated facial recognition was pioneered in the 1960s by Woody Bledsoe, Helen Chan Wolf, and Charles Bisson. Their early system required a human operator to manually mark facial coordinates on a graphics tablet before a computer could process the data. Takeo Kanade publicly demonstrated the first fully automated face-matching system in 1970 and published the first detailed book on facial recognition technology in 1977.
Accuracy varies widely across populations and contexts. Algorithms evaluated in 2006 were 100 times more accurate than those from 1995, and Facebook's DeepFace system reported 97% accuracy on standardized tests. However, the Delhi Police reported its system had an accuracy rate of 2% in 2018, and a 2018 report by Big Brother Watch found UK police live systems were up to 98% inaccurate.
Studies have found significant racial bias. A 2018 study by Joy Buolamwini and Timnit Gebru found that the error rate for gender recognition among women of color in leading commercial systems ranged from 23.8% to 36%, compared to 0.0% to 1.6% for lighter-skinned men. Common image compression methods such as JPEG chroma subsampling have also been found to disproportionately degrade performance for darker-skinned individuals.
Meta Platforms shut down Facebook's facial recognition system in 2021 and deleted the face-scan data of more than one billion users in response to growing societal concerns about privacy and misuse of biometric data. IBM also stopped offering facial recognition technology around the same time for similar reasons.
The use of facial recognition systems has been banned in several cities in the United States following controversy over privacy violations, inaccurate identifications, and racial profiling concerns. In the UK, the Court of Appeal ruled in August 2020 that South Wales Police's use of facial recognition in 2017 and 2018 violated human rights due to an insufficient legal framework and lack of proportionality.