Looking to possibly replace Machinebox/Facebox, looking for advice

Hey guys!Hope you are all doing amazingly well today

I am currently looking to replace Machinebox/Facebox as they can only be taught a hundred references, and know little about machine learning, but am always keen to learn!

I am currently using node-red for this, with a flow that POSTs the facebox server and then analyses its response. All of this is then routed through Home-Assistant for Notifications, Speakers announcement and light coloring!

The ideal software would be something that is; Fast ( This VM has 12 cores, to keep the Inference time to a minimum ), POSTable, Scalable ( No limit on how many faces it can learn),Runs on Ubuntu and if possible, GPU Accelerated ( Got a 560TI with some Cuda cores laying around ).

Facebox also deletes the reference points at each restart, which is very annoying. So if something can save those, it’s even better!

Any idea or comment is appreciated!