[Support]: Home Assistant Cameras Unavailable

[kceleslie]

Describe the problem you are having

I've got frigate and home assistant installed on separate servers. I've gone through the steps here to install and configure HA integration. Everything seems to work except for the cameras. They all (except for birdseye) show up unavailable. When i click on them i should get the RTSP stream but i dont, except for birdseye.

I'm able to use VLC to browse to the RTSP steam, rtsp://frigate:8554/ftdoor manually. I've configured the URL in HA for rtsp to be rtsp://frigate:8554/{{ name|lower }}.

I'm not seeing anything in the HA or frigate logs to help debug this issue. What am I missing?

Version

0.13.0

Frigate config file

mqtt:
  host: mqtt.xxxxx.us
birdseye:
  enabled: true
  mode: continuous
  restream: true
go2rtc:
  streams:
    ftdoor:
      - rtsp://xxxxxxxxxxx/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif
    garage:
      - rtsp://xxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif
    backyard:
      - rtsp://xxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif  
cameras:
  ftdoor:
    ffmpeg:
      inputs:
        - path: rtsp://xxxxxxxx:554/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif
          roles:
            - detect
        - path: rtsp://xxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=0&unicast=true&proto=Onvif
          roles:
            - record
            - rtmp
    detect:
      width: 640
      height: 480
    record:
      enabled: True
      retain:
        days: 10
        mode: all 
    live:
      stream_name: ftdoor       
  garage:
    ffmpeg:
      inputs:
        - path: rtsp://xxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif
          roles:
            - detect
        - path: rtsp://xxxxxxxxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=0&unicast=true&proto=Onvif
          roles:
            - record
            - rtmp
    detect:
      width: 640
      height: 480
    record:
      enabled: True
      retain:
        days: 10
        mode: all
      events:
        retain:
          default: 14
          mode: active_objects
    zones:
      garageStreet:
        coordinates: 218,132,264,113,326,96,381,86,403,77,427,70,501,57,503,41,498,31,473,27,461,26,440,29,422,30,398,33,379,36,355,38,327,39,311,44,282,49,221,62,172,75,120,91,63,109,0,138,0,180,0,143,0,229
      garageNeighborsAcross:
        coordinates: 0,140,90,102,150,80,220,58,320,41,407,32,520,22,525,0,394,0,250,0,0,0
      garageNeighbors:
        coordinates: 562,121,575,41,537,38,437,75
    live:
      stream_name: garage
  backyard:
    ffmpeg:
      inputs:
        - path: rtsp://xxxxxxxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=1&unicast=true&proto=Onvif
          roles:
            - detect
        - path: rtsp://xxxxxxxxxxxxxxxxxxxxx:554/cam/realmonitor?channel=1&subtype=0&unicast=true&proto=Onvif
          roles:
            - record
            - rtmp
    detect:
      width: 640
      height: 480
    record:
      enabled: True
      retain:
        days: 10
        mode: all
      events:
        retain:
          default: 14
          mode: active_objects
    live:
      stream_name: backyard
objects:
  filters:
    person:
      mask:
        - 242,120,242,84,254,82,257,119
        - 1081,194,1096,148,1119,152,1100,203
        - 222,34,233,32,234,53,222,54
  track:
    - person
    - car
    - dog
    - bicycle
    - bus
    - horse
detectors:
  tensorrt:
    type: tensorrt
    model:
      path: /config/model_cache/tensorrt/yolov7-320.trt
record:
  enabled: True
  retain:
    days: 3
    mode: all
  events:
    retain:
      default: 15
      mode: motion
ffmpeg:
  hwaccel_args: preset-jetson-h264
model:
  input_tensor: nchw                              # yolo models are nchw
  width: 320                                      # must match model size
  height: 320                                     # must match model size

Relevant log output

n/a

FFprobe output from your camera

Not sure how to do this

Frigate stats

No response

Operating system

Other Linux

Install method

Docker Compose

Object Detector

TensorRT

Network connection

Wired

Camera make and model

Amcrest

Any other information that may be helpful

No response

[NickM-27]

input_pixel_format: rgb is missing from the model definition

[kceleslie]

Quick search didnt find that in the docs, but i added it and reloaded everything. Still same results

[kceleslie]

Ah ok i see it.
Not sure what you are looking for but here is a screenshot

[NickM-27]

Looking for a screenshot of the whole thing. But this shows the issue anyway, the inference speed of the detector is very slow.

[kceleslie]

that would cause the video feeds to not show up in home assistant?

[NickM-27]

It would cause the cameras to be marked as unavailable because not enough frames are able to be processed so it falls behind.

[NickM-27]

I'd suggest trying a tiny model