[Detector Support]: Cannot detect CATS

[LuckyTitan]

Describe the problem you are having

Current Situation:
frigate running as a container on unraid. Detection of person working fine. Detecting cats does not work at all. Neither at night nor on days. I use a coral tpu and the config below. Happy to provide more details if needed. Help appreciated.

Version

version: 0.15-1

Frigate config file

mqtt:
  enabled: true
  host: 192.168.0.22
  user: mosquitto
  password: Xxxxxxxxxxxxxxxxx
  topic_prefix: frigate
  # Optional: client id (default: shown below)
  # WARNING: must be unique if you are running multiple instances
  client_id: frigate
  # Optional: interval in seconds for publishing stats (default: shown below)
  # stats_interval: 1

database:
  path: /config/frigate.db


detectors:
  coral:
    type: edgetpu
    device: usb

birdseye:
  enabled: true
  restream: true
  mode: continuous

ffmpeg:
  hwaccel_args: preset-vaapi


review:
  alerts:
    labels:
      - person
  detections:
    labels:
      - cat
      - car



go2rtc:
  streams:
    Dachboden:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.72:554/h264Preview_01_main
    Dachboden_sub:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.72:554/h264Preview_01_sub

    Garten:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.73:554/h264Preview_01_main
    Garten_sub:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.73:554/h264Preview_01_sub

    Technikraum:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.70:554/h264Preview_01_main
    Technikraum_sub:
      - rtsp://admin:UxaX95c39mKnzVHb@192.168.0.70:554/h264Preview_01_sub


cameras:
  Dachboden_CAM:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-aac
      inputs:
        - path: rtsp://127.0.0.1:8554/Dachboden?video=copy&audio=aac
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/Dachboden_sub?video=copy
          input_args: preset-rtsp-restream
          roles:
            - detect
    detect:
      width: 640
      height: 360
      fps: 10
    objects:
      track:
        - cat
        - person
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      retain:
        default: 30
    record:
      enabled: true
      retain:
        days: 7
        mode: motion
      alerts:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
      detections:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
    ui:
      order: 3


  Garten_CAM:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-aac
      inputs:
        - path: rtsp://127.0.0.1:8554/Garten?video=copy&audio=aac
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/Garten_sub?video=copy
          input_args: preset-rtsp-restream
          roles:
            - detect
    detect:
      width: 640
      height: 360
      fps: 10
    objects:
      track:
        - cat
        - person
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      retain:
        default: 30
    record:
      enabled: true
      retain:
        days: 7
        mode: motion
      alerts:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
      detections:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
    ui:
      order: 1


    motion:
      threshold: 20
      contour_area: 10
      improve_contrast: true
  Technikraum_CAM:
    ffmpeg:
      output_args:
        record: preset-record-generic-audio-aac
      inputs:
        - path: rtsp://127.0.0.1:8554/Technikraum
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/Technikraum_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
    detect:
      width: 640
      height: 360
      fps: 10
    objects:
      track:
        - cat
    snapshots:
      enabled: true
      timestamp: true
      bounding_box: true
      retain:
        default: 30
    record:
      enabled: true
      retain:
        days: 7
        mode: motion
      alerts:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
      detections:
        retain:
          days: 14
        pre_capture: 10
        post_capture: 10
    ui:
      order: 2
version: 0.15-1

docker-compose file or Docker CLI command

docker run
  -d
  --name='frigate'
  --net='dockerhome'
  --pids-limit 2048
  -e TZ="Europe/Berlin"
  -e HOST_OS="Unraid"
  -e HOST_HOSTNAME="Tower"
  -e HOST_CONTAINERNAME="frigate"
  -e 'FRIGATE_RTSP_PASSWORD'='LQwNyVJJ5QnRKBPi02'
  -e 'LIBVA_DRIVER_NAME'='radeonsi'
  -l net.unraid.docker.managed=dockerman
  -l net.unraid.docker.webui='http://[IP]:[PORT:5000]'
  -l net.unraid.docker.icon='https://raw.githubusercontent.com/yayitazale/unraid-templates/main/frigate.png'
  -p '5000:5000/tcp'
  -p '8554:8554/tcp'
  -p '8555:8555/tcp'
  -p '8555:8555/udp'
  -p '1984:1984/tcp'
  -v '/mnt/user/appdata/frigate':'/config':'rw'
  -v '/mnt/user/Surveillance/frigate/':'/media/frigate':'rw'
  -v '/etc/localtime':'/etc/localtime':'rw'
  --device='/dev/bus/usb'
  --device='/dev/dri/renderD128'
  --shm-size=256mb
  --mount type=tmpfs,target=/tmp/cache,tmpfs-size=1000000000
  --restart unless-stopped 'ghcr.io/blakeblackshear/frigate:stable'
ce41298a157cf0ed619088227af369aa85bed0c8382bdb55af0a42115b1dee60

The command finished successfully!

Relevant Frigate log output

2025-02-11 21:30:11.632989703  [INFO] Preparing Frigate...
2025-02-11 21:30:13.773798159  [INFO] Starting Frigate...
2025-02-11 21:30:21.707419085  [2025-02-11 21:30:21] frigate.util.config            INFO    : Checking if frigate config needs migration...
2025-02-11 21:30:21.743711780  [2025-02-11 21:30:21] frigate.util.config            INFO    : frigate config does not need migration...
2025-02-11 21:30:21.817998631  [2025-02-11 21:30:21] frigate.app                    INFO    : Starting Frigate (0.15.0-cea210d)
2025-02-11 21:30:21.835907106  [2025-02-11 21:30:21] peewee_migrate.logs            INFO    : Starting migrations
2025-02-11 21:30:21.836174218  [2025-02-11 21:30:21] peewee_migrate.logs            INFO    : There is nothing to migrate
2025-02-11 21:30:21.869244270  [2025-02-11 21:30:21] frigate.app                    INFO    : Recording process started: 386
2025-02-11 21:30:21.869619245  [2025-02-11 21:30:21] frigate.app                    INFO    : Review process started: 395
2025-02-11 21:30:21.872633767  [2025-02-11 21:30:21] frigate.app                    INFO    : go2rtc process pid: 98
2025-02-11 21:30:21.891379758  [2025-02-11 21:30:21] detector.coral                 INFO    : Starting detection process: 402
2025-02-11 21:30:21.909066295  [2025-02-11 21:30:21] frigate.app                    INFO    : Output process started: 418
2025-02-11 21:30:21.942892890  [2025-02-11 21:30:21] frigate.detectors.plugins.edgetpu_tfl INFO    : Attempting to load TPU as usb
2025-02-11 21:30:24.547121046  [2025-02-11 21:30:21] frigate.app                    INFO    : Camera processor started for Dachboden_CAM: 435
2025-02-11 21:30:24.547186049  [2025-02-11 21:30:21] frigate.app                    INFO    : Camera processor started for Garten_CAM: 436
2025-02-11 21:30:24.547432382  [2025-02-11 21:30:21] frigate.app                    INFO    : Camera processor started for Technikraum_CAM: 437
2025-02-11 21:30:24.547446168  [2025-02-11 21:30:21] frigate.app                    INFO    : Capture process started for Dachboden_CAM: 445
2025-02-11 21:30:24.547473229  [2025-02-11 21:30:22] frigate.app                    INFO    : Capture process started for Garten_CAM: 465
2025-02-11 21:30:24.547498446  [2025-02-11 21:30:22] frigate.app                    INFO    : Capture process started for Technikraum_CAM: 471
2025-02-11 21:30:24.547522031  [2025-02-11 21:30:22] frigate.api.fastapi_app        INFO    : Starting FastAPI app
2025-02-11 21:30:24.547554792  [2025-02-11 21:30:22] frigate.api.fastapi_app        INFO    : FastAPI started
2025-02-11 21:30:24.550911077  [2025-02-11 21:30:24] frigate.detectors.plugins.edgetpu_tfl INFO    : TPU found

Install method

Docker Compose

Object Detector

Coral

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[hawkeye217]

There are a few factors that may cause this.

1. Your motion detection may not be sensitive enough. Try tuning your motion detection. https://docs.frigate.video/configuration/motion_detection

2. The default free model that ships with Frigate is based on the COCO dataset and is not trained on security camera images. As a result, you may have to lower your threshold or min_score for cat. https://docs.frigate.video/configuration/object_filters

[LuckyTitan]

Thx for the quick reply @hawkeye217 i tried your tip no.1 just now. Different values and a test with my cat did unfortunately not help. Will give the object_filters a try now.

[LuckyTitan]

Good morning (at least where i am from), i tried and set a few different object_filters. The last ones i tried were: filters:
cat:
min_area: 2000 # Adjust depending on the size and distance of the object in view
max_area: 10000 # Limit larger object detections
min_score: 0.6 # Confidence score for detection
threshold: 0.7 # Higher threshold to avoid false positives
min_ratio: 0.4 # Cat's body tends to be longer than it is tall
max_ratio: 2.0

My bigger cat got detected once with these settings, but as a person. Motion boxes in red clearly see both cats moving, but nothing else happens as already described.

[hawkeye217]

I would remove min_area, max_area, min_ratio and max_ratio for now, and lower your min_score and threshold. Try this:

min_score: 0.5
threshold: 0.6

[RealRoeboeter]

Hallo anderer Katzenhalter,
ich hatte dasselbe Problem, war verzweifelt und habe ewig Einstellungen geändert.
Trotzdem wurden fast nie Katzen erkannt.
Die Lösung war die Anmeldung bei Frigate+ und das Trainieren eines eigenen model mit meinen Bildern. Für mein erstes Modell habe ich 200 Bilder an Frigate+ geschickt, und endlich werden meine zwei Katzen bei jeder Bewegung erkannt.
Ich habe dann sogar drei Monate lang vergessen, dass ich noch weitere models trainieren kann, weil es einfach so gut funktioniert hat. Vor Kurzem habe ich mein zweites model mit 467 Bildern und 684 Labels trainieren lassen, und jetzt wird alles Tag und Nacht zuverlässig erkannt.
Ich bin wirklich begeistert – Frigate+ kann ich nur empfehlen!

[LuckyTitan]

Hi anderer Katzehalter und vielen Dank für die ausführliche Antwort. Ich nehme an, du hast die Anleitung auf der Website für die Bilder befolgt und sonst nichts weiter beachtet. Hast du dazu noch Tips für mich? Werde mich dann gleich mal an Frigate+ wagen bzw. registrieren.

[RealRoeboeter]

Genau ich hab mich eigentlich an die Anleitung gehalten . Wenn ich aber meine beigefügte Label Übersicht anschaue waren meine übermittelten Bilder sehr Katzenlastig :) Da meine eine Katze ab und zu als Hund erkannt wurde (geht gar nicht!) habe ich noch angefangen Hunde mit auf den Bildern zu makieren.

Viel Spaß bei das ist eine Katze und das ist keine Katze markieren!