I used to capture screens from my Pikrellcam setup on a second Pi then run them through image recognition using a Coral USB Accelerator sending a notification to my Google Home via Node-Red when an person was detected and waking a screen showing the stream.
It worked well but was very complex and fragile especially as I could only get the image recognition working in a virtual environment.
Anyway I wondered if anyone had done anything similar or can think of a better way to make use of the USB Accelerator?
I used to capture screens from my Pikrellcam setup on a second Pi then run them through image recognition using a Coral USB Accelerator sending a notification to my Google Home via Node-Red when an person was detected and waking a screen showing the stream. It worked well but was very complex and fragile especially as I could only get the image recognition working in a virtual environment. Anyway I wondered if anyone had done anything similar or can think of a better way to make use of the USB Accelerator?