Weird 4k camera issue

[parkroyk]

So all my non-4k cameras are working, objects being detected, recording clips, etc.

The only camera that's currently not working is my 4k camera. For some reason, its snapshot/stream looks much taller than 1.78 aspect (3840x2160); more like 4:3 aspect. I am suspecting perhaps the object detector can't recognize person or car (or anything for that matter) as things look stretched vertically?

I tried setting the width & height to 3840x2160 but didn't make a difference. What else can I try?

ffmpeg:
  camera:
    1_camera_4:
      name: Front (cam4)
      host: !secret camera_4_ip
      port: 554
      path: /live/main
      username: !secret camera_4_username
      password: !secret camera_4_password
      codec: hevc_cuvid
      audio_codec: aac
      width: 3840
      height: 2160
      fps: 15
      substream:
        path: /live/sub
        port: 554
        codec: hevc_cuvid
        audio_codec: aac
        width: 704
        height: 576
        fps: 15
      recorder:
        lookback: 2
        idle_timeout: 3
        retain: 7
        filename_pattern: cam4_%Y%m%d_%H%M%S
        segments_folder: /tmp
        thumbnail:
          filename_pattern: cam4_%Y%m%d_%H%M%S

[parkroyk]

As it turns out, the weird aspect ratio was due to the sub-stream. When I removed it altogether, it shows the video in 1.78 ratio.

So I was hoping that would address the non-detect/recording issue, but it didn't. I still can't get it to trigger, when there's plenty of activity front of that camera. How could this happen? If the video stream can be seen, shouldn't the detector/recorder behave the same as other cameras? I'm really puzzled here.

Where's the location of the log file? I'll further play around with the camera settings and see if it makes any difference. Thanks.

[roflcoopter]

If you configure a substream, it will be used for object detection and other computer vision activities, and the mainstream will be used for recordings. In this way you get a nice balance of good performance of the object detector/motion detector etc while recordings will be of high quality.
The aspect ratio should not be a problem.

My guess is that your motion detector is not sensitive enough, best way to find out is to enable debug logging:

logging:
  default_level: debug

If you then do activities in front of the camera that should trigger recorder you can see how much motion is registered and you can change your configuration accordingly

You can find the logs by running docker logs -f viseron.
Its a bit embarrassing but there is no log file currently, its a work in progress

[parkroyk]

Thanks for a quick response. Where in the config should I put that default_level: debug at? At the top? ffmpeg? at the camera level? detector level? I looked through the documentation but didn't find it anywhere. Please clarify - thanks!

…

On Thu, May 4, 2023 at 12:18 AM Jesper ***@***.***> wrote: If you configure a substream, it will be used for object detection and other computer vision activities, and the mainstream will be used for recordings. In this way you get a nice balance of good performance of the object detector/motion detector etc while recordings will be of high quality. The aspect ratio should not be a problem. My guess is that your motion detector is not sensitive enough, best way to find out is to enable debug logging: logging: default_level: debug If you then do activities in front of the camera that should trigger recorder you can see how much motion is registered and you can change your configuration accordingly You can find the logs by running docker logs -f viseron. Its a bit embarrassing but there is no log file currently, its a work in progress — Reply to this email directly, view it on GitHub <#590 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACG6A4FCWKEMB7SBAXPLPRTXENJ5ZANCNFSM6AAAAAAXVC63F4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[parkroyk]

Ok, so I noticed I'm seeing this error from the problematic 4k camera:

[ERROR ] [viseron.components.ffmpeg.recorder.1_camera_4] - [hevc @ 0x5563a67c2640] log2_parallel_merge_level_minus2 out of range: -1

So I figured perhaps it's the camera's HEVC encoder settings, and it finally works in H264 mode. For some reason this camera in H.265 output works fine with Blue Iris. I'm going to reach out to the camera vendor and see if there's a firmware update that potentially fixes this. And/or I'll look into FFMPEG settings see if there are flags I can use to get it to work.

Sorry this turned out to be an issue with my camera, not Viseron itself!

[roflcoopter]

Ahh okay, nice find!

Regarding the logs for future reference, logger is its own component so you put it at the root of your config.yaml
https://viseron.netlify.app/components-explorer/components/logger

[parkroyk]

Ohh I didn't even notice the Logger component until now, as I have been looking at things under Domains. It would be helpful to newbies if you create a section System Domain and list Logger and Webserver, so all components are mapped to domains. Thanks again for your prompt response!

…

On Thu, May 4, 2023 at 1:49 PM Jesper ***@***.***> wrote: Ahh okay, nice find! Regarding the logs for future reference, logger is its own component so you put it at the root of your config.yaml https://viseron.netlify.app/components-explorer/components/logger — Reply to this email directly, view it on GitHub <#590 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACG6A4GPQR4GLIUA55T4VXDXEQI35ANCNFSM6AAAAAAXVC63F4> . You are receiving this because you authored the thread.Message ID: ***@***.***>