Hi Frederick,
I posted earlier but it was removed for being to vague and uninquisitive.
This is my second approach and by no means am I promoting my suggestion, I only have experience with their pricing and deployment in which I will leave to you to engage and purpose on your own will.
In regards to integrations: Milestone & Genetec have been "integrated" however if you were to consult the manufacturer I will link at the end they can describe how their client versus a VMS client is utilized.
In regards to API/SDK integration, said below company utilizes Linux, Linux containers LXC/Docker. MongoDB NoSQL file system, j.son file apparatus all which are based on application(fulcrum AI) and client functionality and maintenance. This said, those systems needing web provisioned access as an API/SDK with a MongoDB is agnostic in nature and common amongst todays current deployed curriculums.
In regards to the small field of view. I will state that all analytic features ARE dependent on FOV, distance to subject, mounting elevations and resolutions. Those should be the defacto baselines analyzed and proofed prior to settling on any AI subject for software rendering. In other words, the installation may need to tailor to AI applications as some lobby, or entrance type camera FOVs are not accurate for the AI algorithm performance. No one wants to see bald people analytics over time.
This is my non promotional suggestion in which you will have to click the link to reach out and engage your own research, possibilities, references and of course pricing matrix. This is my second attempt to post the link, I hope the forum moderators see that I am not promoting this company only suggesting as that is what you have requested "Any suggestions for face recognigiton…" I am suggesting this company because all of the predecessor analytic companies I have experience in are outdated and not worth the mention. Here you go:
https://vintra.io/
Good Luck!
(Rob K ,,,,don't do it man...….)