Accuracy is heavily dependent on how large the populations you’re dealing with. If you’re comparing footage from surveillance camera vs a database of 100+ mill

Accuracy is heavily dependent on how large the populations you’re dealing with. If you’re comparing footage from surveillance camera vs a database of 100+ million people that’s tough. Most algorithms are being tested at 10^5 vs 10^6. Worse, unlike false positive with fingerprints the suspects will likely look similar to the cops not just the algorithm. There’s a major incentive to opt out simply as a result of false positives.