Hello everyone,

I have a few questions about Frigate and security cameras, and I thought this would be a good place to ask.

I’m new to security cameras and Frigate, so please excuse any basic questions I might have.

I have a PC at home with an i7-8700 CPU running Proxmox, where I plan to install Frigate in an LXC container for device passthrough.

I came across this Amcrest camera on Amazon: Amcrest IP5M-B1276EW-AI. Since Amcrest is recommended, I assume it should work well, but I’d like to confirm before purchasing. If you have any camera recommendations in the $60 range, I’d appreciate them.

I also read that having dual network interfaces is recommended. My router doesn’t support creating new subnets (I don’t know if that would be a problem), and my PC currently has only one network interface. My initial plan was to get a PoE switch and connect the cameras and the router to it, but would getting a separate PCIe network card, and then connect my PC to the switch instead of the router work for creating a separate, internet-less network?

Lastly, I understand that using a Coral accelerator is highly recommended. I’m deciding between the $25 PCIe version and the $60 USB version. Does Frigate benefit from the more expensive USB Coral, or is the $25 PCIe version sufficient? My motherboard is a Gigabyte B365M DS3H, in case that’s relevant.

Thanks in advance for your help!

  • brownmustardminion@lemmy.ml
    link
    fedilink
    arrow-up
    0
    ·
    4 months ago

    I use the pcie coral and it works fine with plenty of processing to spare although I believe mine coast e closer to $50. I have 6 amcrest PoE cameras. You should just buy a PoE switch and directly connect all cameras to it. Then link that directly to your frigate box and lock down access. Any amcrest camera should work well with frigate. I believe they all support rtsp protocol.

    • om1k@sopuli.xyzOP
      link
      fedilink
      arrow-up
      0
      ·
      4 months ago

      My frigate box is a PC which only has one ethernet port, so I should buy a separate network card in order to do that, right? I was thinking of buying a 2.5gb card anyways.

      • brownmustardminion@lemmy.ml
        link
        fedilink
        arrow-up
        0
        ·
        4 months ago

        Yeah I would suggest buying another NIC. They’re cheap, its good security, and it opens up another port upstream for other uses.

  • schizo@forum.uncomfortable.business
    link
    fedilink
    English
    arrow-up
    0
    ·
    4 months ago

    If you can use the openvino stuff, you can skip the coral.

    I actually saw a performance improvement moving from the coral to the openvino/iGPU, amusingly enough.

    • TagMeInSkipIGotThis@lemmy.nz
      link
      fedilink
      arrow-up
      0
      ·
      4 months ago

      I put a passively cooled GT1030 in mine and its doing the decoding & tensorrt detection for 4x 1080p cameras just fine, based on its current load I expect I could add another 4.

      I couldn’t tell you if that specific Amcrest camera will work, but I have 2 of the recommended ones on the frigate docs (https://www.amazon.com/gp/product/B083G9KT4C/) and they work.

    • om1k@sopuli.xyzOP
      link
      fedilink
      arrow-up
      0
      ·
      4 months ago

      I could do that, but I don’t know if the iGPU of my i7 8700 is good enough. Even then, if it means taking load off my CPU I think it would be worth it.

        • om1k@sopuli.xyzOP
          link
          fedilink
          arrow-up
          0
          ·
          4 months ago

          I might try this then. How do you run openvino, is it included in the docker container, or is it external?

          • schizo@forum.uncomfortable.business
            link
            fedilink
            English
            arrow-up
            0
            ·
            4 months ago

            Just a configuration option for Frigate, https://docs.frigate.video/configuration/object_detectors/

            Other than picking that type, I don’t think I had to make any other configuration changes as I was already passing the iGPU through to the container for hardware acceleration.

            (As a side note, even with openvino, 4 cameras using the hardware decoding, AND jellyfin transcodes, the iGPU basically sits at 5% usage. The openvino stuff is shockingly efficient.)