[Config Support]: Support Frigate+ detection

[sebastian-tischer]

Describe the problem you are having

Hi Frigate Team,

As a huge fan, I now just switched to Frigate+ - both to support you and to train my own model. As of now it works great and I only have a question I can't solve right now.

The following image shows the camera positioning and behind the fence 2-3 cars get parked.

To solve constant recognition I used the object mask as below which helps getting no notifications on parked objects.

BUT: Unfortunately it doesn't recognizes moving cars that get into the red area - both left or right.

I am thinking of moving from global objects>filters to camera specific and then adjust the scores.

Do you have other recommendations or ideas maybe?

Thank you again and warm regards
Sebastian

Version

13.2

Frigate config file

model:
  path: plus:https://0a861bbb1c1d837ca5bcac6427c46a64

mqtt:
  enabled: true
  host: 192.168.2.163
  port: 1883
  user: hass-mqtt
  password: mqtt

detectors:
  coral:
    type: edgetpu
    device: usb

ffmpeg:
  hwaccel_args: preset-nvidia-h265

motion:
  threshold: 15
  improve_contrast: True
  contour_area: 10

detect:
  enabled: True
  width: 1280
  height: 720
  fps: 5
  stationary:
    interval: 50
    threshold: 50

objects:
  track:
    - person
    - cat
    - dog
    - car
  filters:
    person:
      min_score: 0.60
      threshold: 0.75
      min_area: 5000
      min_ratio: 0
      max_ratio: 0.8
    car:
      min_score: 0.60
      threshold: 0.75
#      min_area: 5000
#      min_ratio: 1.25
#      max_ratio: 1.8
    cat:
      min_score: 0.60
      threshold: 0.75
#      min_area: 5000
#      min_ratio: 1.25
#      max_ratio: 1.9
    dog:
      min_score: 0.60
      threshold: 0.75
#      min_area: 5000
#      min_ratio: 1.25
#      max_ratio: 1.9

birdseye:
  enabled: True
  mode: continuous

record:
  enabled: True
  retain:
    mode: motion
  events:
    pre_capture: 30
    post_capture: 30
    objects:
      - person
      - cat
      - dog
      - car
    retain:
      default: 30
      mode: motion
      objects:
        person: 30
        cat: 30
        dog: 30
        car: 30

snapshots:
  enabled: True
  clean_copy: True
  bounding_box: True
  crop: False
  retain:
    default: 30
    objects:
      person: 30
      car: 30
      cat: 30
      dog: 30

cameras:
  Einfahrt:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Einfahrt
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Einfahrt-Detect
          input_args: preset-rtsp-restream
          roles:
            - detect
    objects:
      filters:
        car:
          mask:
            - 325,405,1003,383,1077,457,1280,691,1280,720,0,720,0,434
            - 0,372,0,221,545,144,558,333
  Eingang:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Eingang
          input_args: preset-rtsp-restream

          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Eingang-Detect
          input_args: preset-rtsp-restream

          roles:
            - detect
    objects:
      filters:
        car:
          mask:
            - 1035,319,1173,104,1000,52,888,27,699,0,437,0,420,78
            - 393,720,1280,720,1280,439,761,207,169,334
  Garten:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Garten
          input_args: preset-rtsp-restream

          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Garten-Detect
          input_args: preset-rtsp-restream

          roles:
            - detect
  Haus:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Haus
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Haus-Detect
          input_args: preset-rtsp-restream
          roles:
            - detect
  Garage:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Garage
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Garage-Detect
          input_args: preset-rtsp-restream
          roles:
            - detect
    objects:
      filters:
        car:
          mask:
            - 1280,0,1280,720,257,720,250,0
  Klingel:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-copy
      inputs:
        - path: rtsp:https://127.0.0.1:8554/Klingel
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp:https://127.0.0.1:8554/Klingel-Detect
          input_args: preset-rtsp-restream
          roles:
            - detect
    objects:
      filters:
        car:
          mask:
            - 611,451,920,453,904,382,908,331,611,334
            - 339,481,358,438,327,381,0,339,0,434
go2rtc:
  streams:
    Einfahrt:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/11#video=copy#audio=copy#audio=opus"
    Einfahrt-Detect:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/12"
    Eingang:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/11#video=copy#audio=copy#audio=opus"
    Eingang-Detect:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/12"
    Garten:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/11#video=copy#audio=copy#audio=opus"
    Garten-Detect:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/12"
    Haus:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/11#video=copy#audio=copy#audio=opus"
    Haus-Detect:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/12"
    Garage:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/11#video=copy#audio=copy#audio=opus"
    Garage-Detect:
      - "ffmpeg:rtsp:https://operator-kamera:[email protected]/livestream/12"
    Klingel:
      - "ffmpeg:rtsp:https://ghusmf0002:[email protected]/mpeg/media.amp"
    Klingel-Detect:
      - "ffmpeg:rtsp:https://ghusmf0002:[email protected]/mpeg/media.amp"

Relevant log output

-

Frigate stats

-

Operating system

UNRAID

Install method

HassOS Addon

Object Detector

Coral

Any other information that may be helpful

No response

[NickM-27]

object masks are not the solution here. Notifications can already be tuned to ignore objects that are stationary. This is in general also something that has been greatly improved in the 0.14 version

[sebastian-tischer]

Ok got it :) Would it help to put object filter again until the model is better trained?

[NickM-27]

it will make it not trigger notifications, so if that is desirable then yes you would want to add it back for now. If that car is consistently in that position you can always use a small object mask just in that specific point so cars that drive by are still detected

[sebastian-tischer]

Alright - just made the object mask and it is much better, now no notification.

What I don't understand, is that it states here, that cars that enter and leave are recorded, just stationary objects are "excluded":
https://docs.frigate.video/guides/parked_cars/

Advantages of this approach
Using this approach, you will get two separate events for when a car enters the driveway, parks in the parking zone, and then later > leaves the zone. Using an object mask will ensure that cars parked in the parking zone are not detected and confused with cars > driving by on the street as well.

Or do I get it wrong?

[NickM-27]

this is correct, however in the video showed above the bounding box varied so much that the object was considered moving (so it is not stationary, but the car did not move)

[sebastian-tischer]

Allright - so I will train the model with the car there even more and try to remove it later.

Thanks a lot for the help - made things much more clear 👍