Second stream for channels



  • with the hability to load a second stream, could we make analitics or motion detection on a lower resolution image, it will uses less cpu, so use the main stream for record, and the second for motion detections. Thank you



  • THAT would be sweet. Good idea. 😎



  • Have you tried setting up a main stream and sub stream with motion detection and comparing the system load?

    ffmpeg outputs small jpeg pictures with a resolution and rate defined by you.

    My setup is running with a very low cpu usage on 1920x1080 main streams, i doubt implementing the sub stream would make a significant difference, worth the effort?
    It might even increase the load by ffmpeg processing 2 streams instead of 1?

    I know it was decreasing the cpu load using the substream with Zoneminder, but i believe Shinobi is doing it smarter using ffmpeg to make the "sub stream" for motion detection 🙂



  • @pamo said in Second stream for channels:

    ffmpeg outputs small jpeg pictures with a resolution and rate defined by you.
    For motion detection? Could you tell me where to have ffmpeg kick out a str
    My setup is running with a very low cpu usage on 1920x1080 main streams, i doubt implementing the sub stream would make a significant difference, worth the effort?

    What's your system? I'm using a Mac Mini (old one I had sitting around, Mid 2011 model, 2.5Ghz i5, 8GB ram, Radeon 6630) and Shinobi reports 135% load (iStat reports 115%) for six 2048x1536 streams, 10fps, h264, Variable/Highest, 1024kbps, 80 I-frame (Hikvision cams). It's not bad, but it's not idling either (cores are running around 200 degrees), and I don't have motion detection on as I haven't figured out how to install it on OS X yet, so perhaps I jumped the gun in agreeing that it would be a good idea.



  • @super_kev said in Second stream for channels:

    @pamo said in Second stream for channels:
    ffmpeg outputs small jpeg pictures with a resolution and rate defined by you.

    For motion detection? Could you tell me where to have ffmpeg kick out a str

    I'm just trying to explain the way Shinobi uses ffmpeg to generate jpegs for the motion detector.
    Resolution and rate of the pictures is just set in the Motion setup for each monitor.

    Get ffmpeg kick out a string ??

    What's your system?

    It's currently a cheap self-made desktop
    i3-8100 3,6GHz. 4 cores in the new 8th gen
    2 x 8GB DDR4
    No additional graphics yet.
    Windows 10 Hyper-V: UbuntuServer 16.04

    Motion detection on
    3x1080p 5fps 2048kbps
    2x640x480 5fps HQ setting

    4 CPUs on linux: 5 % load
    Same CPU load in windows

    I use it for movies and personal use too, thats why i run windows on top 🙂



  • @pamo said in Second stream for channels:

    No additional graphics yet.

    Don't forget that there is no hardware acceleration for Linux in Hyper-V.

    Because of that i thought about Proxmox. On the other hand it should not be needed if the cam delivers h.264 and all there is to do is recording the stream to the HDD and send it as is to the viewing devices.

    But that's what i haven't figured out yet in Shinobi:
    https://forum.shinobi.video/topic/115/rtsp-stream-for-viewing



  • @mrglasspoole said in Second stream for channels:

    @pamo said in Second stream for channels:

    No additional graphics yet.

    Don't forget that there is no hardware acceleration for Linux in Hyper-V.

    Not with the RemoteFX function.
    But if dedicating a GPU by PCI Passthrough to the Linux VM is might be possible after all 🙂



  • I was looking into that and it seems not possible with Linux.



  • Even if not useful for motion detection performance this could be useful for storing video at a lower resolution and/or framerate to allow viewing a time-lapse with limited bandwidth without having to reencode.


 

Looks like your connection to Shinobi Forum was lost, please wait while we try to reconnect.