Hi! I have a question about optimizing Tensorflow. I use supermicro rack server with Xeon E3-1220v3 and three cameras in Tensorflow. The machine is way too much for HASS but doesn’t looks good with Tensorflow. Often I got log message:
Updating tensorflow image_processing took longer than the scheduled update interval 0:00:10
Do you have any solutions or Tensorflow models best for recognising person and car using that CPU. The best result for me is scan camera every 1-2 seconds but I know it can be very hard to achieve. Has someone optimised it for CPU like this?
Here is record from htop with my CPU usage:
https://streamable.com/obcfn