So the result is a camera that yells ‘Doggo!’ more or less at random.
Missed a chance for SQUIRREL
Several minute delay? Have you considered running Frigate for your detections?
minutes delay?
Should be seconds or fractions there if with even basic processing, even in x86
What model architecture?
Right? My Reolink detect on device so it’s definitely ms. Even having HA handle screenshots with the delayed feed through AgentDVR would make it under 2s. OP’s NVR is apparently running it through AfterEffects Pro and airbrushing before export
That’s what I’m saying. YOLOv8 nano and small are like… microscopic. you could run them on like, a graphing calculator.
The cameras are connected to a Unifi Dream Machine Pro. Home Assistant is running on a Raspberry Pi 4.
Something must be wrong with your setup. My G5 detects cars, people and animals basically in real time, including home assistant notifications.
¯_(ツ)_/¯ maybe. I assumed the detection happened on the NVR (Dream Machine Pro in my case).
you dropped this \
Nope, it should be the camera itself. The G3s don’t have detection, connected to the same Dream Machine.
