I have started to use claude sonnet 3.5/01 preview etc to work through an app bucket list

https://binreminded.medium.com/my-ai-code-bucket-list-a-work-in-progress-d13f7ed1d30c

But for facial ID I wanted to have known good faces to compare. A list of matches and mismatches and if the mismatches match it's not working correctly Does it exist? How do the police approve a system if there is no metric of matching/mismatching?