'Nodect' without decompression?
Posted: Tue Mar 10, 2020 1:06 am
I use 'modect' on a network camera sub-stream (640x480x15fps) to trigger the main-stream (1920x1080x30fps). Two monitors, one camera. I'm trying to save resources/cpu power. The sub-stream consumes 8% CPU (with motion detection) if it runs alone. Additionally enabling the main-stream with 'nodect' linked to the sub-stream raises cpu to 21%. The mainstream (alone) on 'modect' would consume 25% cpu power.
25% >> 21% is not really a big improvement. The question is why is it consuming +13% cpu at all in 'nodect'? My guess is ZM still decompresses the h264 stream (that is not necessary here).
Is it possible to disable decompression on 'nodect'?
The best approach would be if ZM could optionally reduce data on-the-fly (process only every 2nd/3rd pixel/row) to lower analysis data. It's 62% less CPU power (21%->8%) per camera (in the case above). That's not bad.
25% >> 21% is not really a big improvement. The question is why is it consuming +13% cpu at all in 'nodect'? My guess is ZM still decompresses the h264 stream (that is not necessary here).
Is it possible to disable decompression on 'nodect'?
The best approach would be if ZM could optionally reduce data on-the-fly (process only every 2nd/3rd pixel/row) to lower analysis data. It's 62% less CPU power (21%->8%) per camera (in the case above). That's not bad.