> Turn on face recognition & upload your first face via Face Library → Add Face.
> Train and improve accuracy: New detections appear in Face Library → Train with a confidence score-assign each to a new or existing person to refine future recognition.
I also ran Doublestake and Compreface with Frigate. Found out that it didn't really provide any benefits for me. The default native person detection in Frigate using the TPU is more than adequate. I've seen some interesting stuff people have done using a mix of locally hosted LLM vision model with Home Assistant and Frigate to do image interpretation. Including facial recognition and License plate reader. It's something I want to eventually explore.
> Turn on face recognition & upload your first face via Face Library → Add Face.
> Train and improve accuracy: New detections appear in Face Library → Train with a confidence score-assign each to a new or existing person to refine future recognition.