Slashdot
Follow
How Accurate Are Flock's AI-Powered License-Reading Cameras?
A Los Angeles Police Department audit revealed that Flock ALPR cameras generated a 32.3 percent error rate in false stolen-vehicle alerts. Flock's CEO cited the FBI's outdated NCIC database as a major issue, explaining their "suppression" feature to work around it. He claimed less than one in a million alerts are questioned by officers. However, a Sacramento officer reported Flock cameras sent multiple false alerts for the same vehicle due to misreading license plate characters. Roseville police found Flock's software misread license plates in 71 percent of alerts, with cameras missing vehicles and sending delayed notifications. Roseville's unique camera setup, while intended to protect privacy, may have contributed to lower performance. In Toledo, a misread license plate by a Flock camera led to a severe dog attack on a driver. Flock stated Roseville's performance has improved, a claim disputed by the police department. Flock's product director suggested drivers remove license plate frames, which Roseville declined. Similar character misreads and blurry images were identified as issues. An independent 2021 test by IPVM also found significant inaccuracies in Flock's license plate readers. Flock reportedly blocked IPVM from further testing after their findings.