Good luck.

The major failing points of facial recognition:

1. Requires a very large, and actually racially diverse, dataset to even just function at a basic level since many that are designed with a core Caucasian [or whatever] dataset flat out can't accurate pick up anyone who is black let alone ID them. Humans are bad enough when it comes to distinguishing between various Chinese and Japanese individuals that an AI won't help much there when it's coded with those actual prejudices towards what counts as a notable feature to use for an IDing marker.

2. Even removing hats, glasses, and every other accessory someone may be wearing you can still going to encounter physical features that fuck these up. Tattoos might be manageable but beards do not play nice with these things at all. I had to renew my passport a while back and it failed so many times in the end I had to tick a box that confirmed I was doing everything I should have been as well as include a note saying "My mouth is closed" because my beard is that fucking big probably only a third of my face is actually visible at any one time.

Shitposting thought: I wonder what would happen if you gave a facial recognition tool a large enough dataset that it might end up labelling anyone with a large enough nose as belonging to certain ethic groups 👀