Better face recognition wanted

Hello, not sure if this is the right pace. I use frigate with double take and codepeoject ai, but I getopften wrong faces detected. So faces it learned are added to wrong persons or unknown faces. I have an Apple TV running on network that is way better in detecting. Also if Angabe is wrong I can click a button and say it was wrong to improve detecting.

I tried also deepstack but it is crashing very often, and it looks to be dead.

So I am looking for a better face recognition system because the Apple detection can not be used for anything then sending a message. Something that is more reliable.

Does anyone uses a good system?

I am using an Intel Nuc so having some resources available but not a core i5 or something with modern codecs so I am a bit limited maybe. But could add a coral usb stick like I have for frigate.

1 Like

You could try CompreFace. It can be connected to double take. I use it and I am satisfied with the results. It requires some memory though. I configured my Docker CompreFace VM with 6 GB RAM.
https://github.com/exadel-inc/CompreFace

1 Like

Unfortunately I can not use compreface because my CPU ist missing AVX it looks like. :expressionless: