[Support]: Dog or Pony

[bb61523]

Describe the problem you are having

I'm get a lot of false positives on dog recognition. I track both horse and dog, since there's no specific pony object I believe?
But unfortunately the pony's get tagged as dogs a lot. And for pony's they're actually even not that small.
I thought about increasing the threshold for dog, but as you can see it's high in the 80%, so I'll miss lot of actual dogs if I set it to 90 or higher.
Is this a Yolo thing? I tried with Yolov9 and Yolov26 medium 320. But more or less same results with both.
Object frame size is between 5000-30000pixels, so I don't think it's too small?
Or is this simply limit of the technology that it can't possibly differentiate between a large dog and small horse?

The genAI object description also doesn't question the fact that this is a dog: "The dog grazes peacefully, occasionally lifting its head. It moves slowly across the lawn, appearing content and undisturbed, suggesting no immediate intent to chase or flee, simply enjoying its outdoor space."

Any help or feedback is appreciated. Let me know if you need any further info.

relevant parts of code:
model:
model_type: yolo-generic
path: /config/model_cache/yolov26m-320.onnx
width: 320
height: 320
input_tensor: nchw
input_dtype: float
labelmap_path: /labelmap/coco-80.txt

objects:
track:
- person
- car
- dog
- cat
- horse

HikTuin:
ffmpeg:
inputs:
- path: rtsp://127.0.0.1:8554/HikTuin
roles:
- record
- path: rtsp://127.0.0.1:8554/HikTuin_sub
roles:
- detect
detect:
width: 1280
height: 720
fps: 5

Version

0.17.1-416a9b7

What browser(s) are you using?

No response

Frigate config file

mqtt:
  host: 192.168.100.103
  user: frigate-user
  password: piep

semantic_search:
  enabled: true
  model: jinav1
  model_size: small

#logger:
#  default: info
#  logs:
#    frigate.genai: debug
#    frigate.embeddings: debug
#    frigate.data_processing: debug
#    frigate.review: debug
#    frigate.detectors: debug
#    frigate.data_processing.real_time.face: debug
#    frigate.data_processing.real_time.custom_classification: debug

ffmpeg:
  hwaccel_args:
    - -hwaccel
    - cuda
    - -hwaccel_output_format
    - h264_cuvid
go2rtc:
  api:
    listen: :1984
  webrtc:
    candidates:
      - 192.168.100.102:8555
  streams:
    HikVoordeur: 
      rtsp://frigate:piep@192.168.110.10:554/Streaming/Channels/101
    HikVoordeur_sub: 
      rtsp://frigate:piep@192.168.110.10:554/Streaming/Channels/102

    HikOprit: 
      rtsp://frigate:piep@192.168.110.11:554/Streaming/Channels/101
    HikOprit_sub: 
      rtsp://frigate:piep@192.168.110.11:554/Streaming/Channels/102

    HikTerras: 
      rtsp://frigate:piep@192.168.110.12:554/Streaming/Channels/101
    HikTerras_sub: 
      rtsp://frigate:piep@192.168.110.12:554/Streaming/Channels/102

    HikTuin: 
      rtsp://frigate:piep@192.168.110.13:554/Streaming/Channels/101
    HikTuin_sub: 
      rtsp://frigate:piep@192.168.110.13:554/Streaming/Channels/103

    HikTuinhuis: 
      rtsp://frigate:piep@192.168.110.14:554/Streaming/Channels/101
    HikTuinhuis_sub: 
      rtsp://frigate:piep@192.168.110.14:554/Streaming/Channels/102

    HikChicks: 
      rtsp://frigate:piep@192.168.110.15:554/Streaming/Channels/101
    HikChicks_sub: 
      rtsp://frigate:piep@192.168.110.15:554/Streaming/Channels/102

    HikAchter: 
      rtsp://frigate:piep@192.168.110.16:554/Streaming/Channels/101
    HikAchter_sub: 
      rtsp://frigate:piep@192.168.110.16:554/Streaming/Channels/102

    HikWei: 
      rtsp://frigate:piep@192.168.110.17:554/Streaming/Channels/101
    HikWei_sub: 
      rtsp://frigate:piep@192.168.110.17:554/Streaming/Channels/102

detectors:
  onnx1:
    type: onnx
    device: '0'
    providers:
      - TensorrtExecutionProvider
      - CUDAExecutionProvider
    options:
      trt_engine_cache_enable: true
      trt_engine_cache_path: /config/model_cache
      trt_fp16_enable: true
#  onnx2:
#    type: onnx
#    device: '0'
#    providers:
#      - TensorrtExecutionProvider
#      - CUDAExecutionProvider
#    options:
#      trt_engine_cache_enable: true
#      trt_engine_cache_path: /config/model_cache
#      trt_fp16_enable: true
model:
  model_type: yolo-generic
  path: /config/model_cache/yolov26m-320.onnx
  width: 320
  height: 320
  input_tensor: nchw
  input_dtype: float
  labelmap_path: /labelmap/coco-80.txt

detect:
  enabled: true
  stationary:
    interval: 50
    threshold: 50

face_recognition:
  enabled: true
  model_size: large   # gebruik 'small' als je geen GPU wilt inzetten
  detection_threshold: 0.5   # 0.7 standaard, pas aan indien nodig
  recognition_threshold: 0.85  # hoe zeker Frigate moet zijn voor een match
  min_area: 600              # 500 standaard, minimale pixelgrootte van een gezicht
  blur_confidence_filter: true  # filtert wazige gezichten (standaard aan)

genai:
  provider: ollama
  base_url: http://192.168.100.102:11434
  model: qwen3-vl:8B-Instruct

motion:
  lightning_threshold: 0.5 #standard 0.6
  threshold: 40 #45standard 30, increased for less motion blur during night
  contour_area: 25 #30standard 10, increased for less motion blur during night


review:
  alerts:
    labels:
      - person
      - car
      - dog
  genai:
    enabled: false
    image_source: preview # Options: "preview" (default) or "recordings" for better quality
    additional_concerns:
      - Deer or wild pigs in the garden

ui:
  time_format: 24hour

objects:
  track:
    - person
    - car
    - dog
    - cat
    - horse
  filters:
    person:
      min_score: 0.70
      threshold: 0.70
  genai:
    enabled: true
    use_snapshot: false
    prompt: "Analyze the {label} in these images from the {camera} security camera. Focus on the actions, behavior, and potential intent of the {label}, rather than just describing its appearance. Be concise, Limit your response to a maximum of 35 words."
    object_prompts:
      person: "Examine the main person in these images. What are they doing and what might their actions suggest about their intent (e.g., approaching a door, leaving an area, standing still)? Do not describe the surroundings or static details. Be concise, Limit your response to a maximum of 35 words."
      car: "Observe the primary vehicle in these images. Focus on its movement, direction, or purpose (e.g., parking, approaching, circling). If it's a delivery vehicle, mention the company. Be concise, Limit your response to a maximum of 35 words."

classification:
  custom:
    Obi:
      enabled: true
      name: Obi
      threshold: 0.8
      object_config:
        objects:
          - dog
        classification_type: sub_label

  bird:
    enabled: false
database:
  path: /config/frigate.db

record:
  enabled: true
  continuous:
    days: 15
  motion:
    days: 30
snapshots:
  enabled: true
  timestamp: true
  bounding_box: true
  retain:
    default: 60

cameras:
  HikVoordeur:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikVoordeur
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikVoordeur_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    zones:
      Voordeur:
        coordinates: 
          0,1,0.091,0.545,0.193,0.392,0.323,0.643,0.488,0.477,0.695,0.375,0.707,0.469,0.851,0.006,0.995,0.008,1,1
        inertia: 3
        loitering_time: 0
    motion:
      mask: 
        0.068,0.071,0.11,0.005,0.236,-0.002,0.294,0.272,0.37,0.539,0.382,0.591,0.314,0.668
  HikOprit:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikOprit
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikOprit_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 0.996,0.384,0.637,0.257,0.614,0.127,0,0.14,0.003,0.002,0.996,0.002
      improve_contrast: false
    objects:
      filters:
        person:
          mask: 
            0.275,0.651,0.276,0.413,0.307,0.409,0.352,0.41,0.36,0.649,0.319,0.649,0.308,0.648
        car:
          mask: 
            0,0.144,0.413,0.135,0.423,0.303,0.361,0.389,0.436,0.477,0.518,0.379,0.592,0.32,0.617,0.315,0.654,0.333,0.694,0.338,0.722,0.338,1,0.424,1,1,0,1
  HikTerras:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikTerras
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikTerras_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask:
        - 0.171,0.996,0.003,0.487,0,0.999
        - 0.714,0.998,0.822,0.415,0.91,0.004,1,0.001,0.997,0.995
      improve_contrast: false
  HikTuin:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikTuin
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikTuin_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 0.998,0.24,0.002,0.267,0.002,0.005,0.998,0.005

  HikTuinhuis:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikTuinhuis
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikTuinhuis_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 0.003,0.304,0.366,0.302,0.71,0.305,1,0.322,1,0,0.002,-0.001
  HikChicks:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikChicks
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikChicks_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 
        0.999,0.232,0.385,0.312,0.378,0.383,0.191,0.387,0.001,0.728,0.001,0,0.994,0.002
  HikAchter:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikAchter
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikAchter_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 
        0,0.255,0.448,0.201,0.774,0.192,0.876,0.207,0.856,0.739,0.931,0.775,0.998,0,0.003,0.003
    objects:
      filters:
        car:
          mask: 0.259,0.447,0.328,0.627,0.516,0.486,0.4,0.347,0.317,0.37
  HikWei:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/HikWei
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/HikWei_sub
          roles:
            - detect
    detect:
      width: 1280
      height: 720
      fps: 5

    motion:
      mask: 0.996,0.42,0.283,0.207,0.383,0.002,0.998,0.002
      improve_contrast: false
version: 0.17-0
lpr:
  enabled: false

Relevant Frigate log output

na

Relevant go2rtc log output

na

FFprobe output from your camera

na

Frigate stats

{"cameras":{"HikVoordeur":{"camera_fps":5.0,"process_fps":5.0,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1153,"capture_pid":1214,"ffmpeg_pid":1267,"audio_rms":0,"audio_dBFS":0},"HikOprit":{"camera_fps":5.1,"process_fps":5.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1280,"capture_pid":1335,"ffmpeg_pid":1390,"audio_rms":0,"audio_dBFS":0},"HikTerras":{"camera_fps":5.2,"process_fps":5.2,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1403,"capture_pid":1460,"ffmpeg_pid":1513,"audio_rms":0,"audio_dBFS":0},"HikTuin":{"camera_fps":5.1,"process_fps":4.9,"skipped_fps":0.2,"detection_fps":3.1,"detection_enabled":true,"pid":1526,"capture_pid":1581,"ffmpeg_pid":1635,"audio_rms":0,"audio_dBFS":0},"HikTuinhuis":{"camera_fps":5.1,"process_fps":5.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1649,"capture_pid":1704,"ffmpeg_pid":1762,"audio_rms":0,"audio_dBFS":0},"HikChicks":{"camera_fps":5.0,"process_fps":5.0,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1775,"capture_pid":1832,"ffmpeg_pid":1886,"audio_rms":0,"audio_dBFS":0},"HikAchter":{"camera_fps":5.0,"process_fps":5.0,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":1899,"capture_pid":1956,"ffmpeg_pid":2011,"audio_rms":0,"audio_dBFS":0},"HikWei":{"camera_fps":5.0,"process_fps":5.0,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":2024,"capture_pid":2080,"ffmpeg_pid":2128,"audio_rms":0,"audio_dBFS":0}},"detectors":{"onnx1":{"inference_speed":10.03,"detection_start":0.0,"pid":1021}},"camera_fps":40.5,"process_fps":40.3,"skipped_fps":0.2,"detection_fps":3.1,"embeddings":{"image_embedding_speed":107.41,"image_embedding":0.0,"text_embedding_speed":113.12,"text_embedding":0.0,"face_recognition_speed":167.66,"face_recognition":0.0,"object_description_speed":3433.44,"object_description_events_per_second":0.0,"Obi_classification_speed":3.21,"Obi_classification_events_per_second":0.0},"gpu_usages":{"NVIDIA RTX PRO 2000 Blackwell":{"gpu":"1.0%","mem":"61.92%","enc":"0.0%","dec":"3.0%"}},"cpu_usages":{"frigate.full_system":{"cpu":"7.6","mem":"33.0"},"1":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/package/admin/s6/command/s6-svscan -d4 -- /run/service"},"61":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6-linux-init-shutdownd"},"64":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/package/admin/s6-linux-init/command/s6-linux-init-shutdownd -d3 -c /run/s6/basedir -g 3000 -C -B"},"71":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise certsync"},"72":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise nginx-log"},"73":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise certsync-log"},"74":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise frigate-log"},"75":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6rc-oneshot-runner"},"76":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc-log"},"77":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6rc-fdholder"},"78":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc-healthcheck"},"79":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise nginx"},"80":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise frigate"},"81":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc"},"92":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/package/admin/s6-2.13.2.0/command/s6-fdholderd -1 -n 256 -i data/rules"},"93":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/package/admin/s6/command/s6-ipcserverd -1 -- /package/admin/s6/command/s6-ipcserver-access -v0 -E -l0 -i data/rules -- /package/admin/s6/command/s6-sudod -t 30000 -- /package/admin/s6-rc/command/s6-rc-oneshot-run -l ../.. --"},"140":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/nginx"},"142":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/go2rtc"},"143":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/frigate"},"144":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/certsync"},"172":{"cpu":"17.5","cpu_average":"17","mem":"0.2","cmdline":"/usr/local/go2rtc/bin/go2rtc -config=/config/go2rtc_homekit.yml -config=/dev/shm/go2rtc.yaml"},"180":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"bash ./run.user go2rtc-healthcheck"},"181":{"cpu":"4.4","cpu_average":"4","mem":"3.6","cmdline":"python3 -u -m frigate"},"187":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: master process nginx"},"254":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"255":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"256":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"257":{"cpu":"0.1","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"258":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: cache manager process"},"429":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"bash ./run.user certsync"},"477":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/bin/python3 -Wignore:::numpy.core.getlimits -c from multiprocessing.resource_tracker import main;main(5)"},"478":{"cpu":"0.0","cpu_average":"0","mem":"0.7","cmdline":"/usr/bin/python3 -Wignore:::numpy.core.getlimits -c from multiprocessing.forkserver import main; main(5, 7, ['sqlite3', 'numpy', 'cv2', 'peewee', 'zmq', 'ruamel.yaml', 'frigate.camera.maintainer'], **{'sys_path': ['/opt/frigate', '/usr/lib/python311.zip', '/usr/lib/python3.11', '/usr/lib/python3.11/lib-dynload', '/usr/local/lib/python3.11/dist-packages', '/usr/lib/python3/dist-packages']})"},"508":{"cpu":"2.2","cpu_average":"2","mem":"0.4","cmdline":"/usr/bin/python3 -Wignore:::numpy.core.getlimits -c from multiprocessing.forkserver import main; main(5, 7, ['sqlite3', 'numpy', 'cv2', 'peewee', 'zmq', 'ruamel.yaml', 'frigate.camera.maintainer'], **{'sys_path': ['/opt/frigate', '/usr/lib/python311.zip', '/usr/lib/python3.11', '/usr/lib/python3.11/lib-dynload', '/usr/local/lib/python3.11/dist-packages', '/usr/lib/python3/dist-packages']})"},"960":{"cpu":"9.2","cpu_average":"7","mem":"0.6","cmdline":"frigate.recording_manager                                                                                                                                                                                                                                                                                                                                                                               "},"973":{"cpu":"0.8","cpu_average":"0","mem":"0.5","cmdline":"frigate.review_segment_manager                                                                                                                                                                                                                                                                                                                                                                          "},"983":{"cpu":"1.7","cpu_average":"3","mem":"9.5","cmdline":"frigate.embeddings_manager                                                                                                                                                                                                                                                                                                                                                                              "},"1021":{"cpu":"6.2","cpu_average":"3","mem":"2.7","cmdline":"frigate.detector:onnx1                                                                                                                                                                                                                                                                                                                                                                                  "},"1032":{"cpu":"1.7","cpu_average":"1","mem":"0.6","cmdline":"frigate.output                                                                                                                                                                                                                                                                                                                                                                                          "},"1057":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1062":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1067":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1072":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1077":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1084":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1089":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1094":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1099":{"cpu":"0.0","cpu_average":"0","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -threads 1 -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -i pipe: -threads 1 -f mpegts -s 1280x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"1153":{"cpu":"1.1","cpu_average":"1","mem":"0.5","cmdline":"frigate.process:HikVoordeur                                                                                                                                                                                                                                                                                                                                                                             "},"1214":{"cpu":"0.9","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikVoordeur                                                                                                                                                                                                                                                                                                                                                                             "},"1267":{"cpu":"1.5","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikVoordeur_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1279":{"cpu":"1.6","cpu_average":"1","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikVoordeur -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikVoordeur@%Y%m%d%H%M%S%z.mp4"},"1280":{"cpu":"0.6","cpu_average":"0","mem":"0.5","cmdline":"frigate.process:HikOprit                                                                                                                                                                                                                                                                                                                                                                                "},"1335":{"cpu":"0.8","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikOprit                                                                                                                                                                                                                                                                                                                                                                                "},"1390":{"cpu":"1.4","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikOprit_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1402":{"cpu":"1.1","cpu_average":"2","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikOprit -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikOprit@%Y%m%d%H%M%S%z.mp4"},"1403":{"cpu":"0.6","cpu_average":"0","mem":"0.5","cmdline":"frigate.process:HikTerras                                                                                                                                                                                                                                                                                                                                                                               "},"1460":{"cpu":"0.9","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikTerras                                                                                                                                                                                                                                                                                                                                                                               "},"1513":{"cpu":"1.4","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTerras_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1525":{"cpu":"1.4","cpu_average":"1","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTerras -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikTerras@%Y%m%d%H%M%S%z.mp4"},"1526":{"cpu":"4.3","cpu_average":"1","mem":"0.5","cmdline":"frigate.process:HikTuin                                                                                                                                                                                                                                                                                                                                                                                 "},"1581":{"cpu":"0.9","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikTuin                                                                                                                                                                                                                                                                                                                                                                                 "},"1635":{"cpu":"1.4","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTuin_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1648":{"cpu":"0.4","cpu_average":"0","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTuin -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikTuin@%Y%m%d%H%M%S%z.mp4"},"1649":{"cpu":"1.1","cpu_average":"1","mem":"0.5","cmdline":"frigate.process:HikTuinhuis                                                                                                                                                                                                                                                                                                                                                                             "},"1704":{"cpu":"0.8","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikTuinhuis                                                                                                                                                                                                                                                                                                                                                                             "},"1762":{"cpu":"1.3","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTuinhuis_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1774":{"cpu":"0.5","cpu_average":"0","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikTuinhuis -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikTuinhuis@%Y%m%d%H%M%S%z.mp4"},"1775":{"cpu":"1.1","cpu_average":"1","mem":"0.5","cmdline":"frigate.process:HikChicks                                                                                                                                                                                                                                                                                                                                                                               "},"1832":{"cpu":"0.9","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikChicks                                                                                                                                                                                                                                                                                                                                                                               "},"1886":{"cpu":"1.4","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikChicks_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"1897":{"cpu":"1.4","cpu_average":"1","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikChicks -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikChicks@%Y%m%d%H%M%S%z.mp4"},"1899":{"cpu":"1.1","cpu_average":"1","mem":"0.5","cmdline":"frigate.process:HikAchter                                                                                                                                                                                                                                                                                                                                                                               "},"1956":{"cpu":"0.8","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikAchter                                                                                                                                                                                                                                                                                                                                                                               "},"2011":{"cpu":"1.4","cpu_average":"1","mem":"0.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikAchter_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"2021":{"cpu":"1.3","cpu_average":"1","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikAchter -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikAchter@%Y%m%d%H%M%S%z.mp4"},"2024":{"cpu":"0.8","cpu_average":"0","mem":"0.5","cmdline":"frigate.process:HikWei                                                                                                                                                                                                                                                                                                                                                                                  "},"2080":{"cpu":"0.9","cpu_average":"0","mem":"0.4","cmdline":"frigate.capture:HikWei                                                                                                                                                                                                                                                                                                                                                                                  "},"2128":{"cpu":"1.3","cpu_average":"1","mem":"0.5","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format h264_cuvid -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikWei_sub -r 5 -vf fps=5,scale=1280:720 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"2135":{"cpu":"0.3","cpu_average":"0","mem":"0.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.17.1-416a9b7 -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 10000000 -use_wallclock_as_timestamps 1 -i rtsp://127.0.0.1:8554/HikWei -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/HikWei@%Y%m%d%H%M%S%z.mp4"},"959191":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"sleep 60"},"959324":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"sleep 30s"}},"service":{"uptime":35810,"version":"0.17.1-416a9b7","latest_version":"0.17.1","storage":{"/media/frigate/recordings":{"total":16873876.0,"used":10207512.0,"free":6665836.0,"mount_type":"nfs4"},"/media/frigate/clips":{"total":16873876.0,"used":10207512.0,"free":6665836.0,"mount_type":"nfs4"},"/tmp/cache":{"total":953.7,"used":32.9,"free":920.8,"mount_type":"tmpfs"},"/dev/shm":{"total":2048.0,"used":531.8,"free":1516.2,"mount_type":"tmpfs","available":1998.0,"camera_frame_size":16.0,"shm_frame_count":50,"min_shm":370}},"temperatures":{},"last_updated":1780164231},"processes":{"recording":{"pid":960},"review_segment":{"pid":973},"go2rtc":{"pid":172},"embeddings":{"pid":983}}}

Install method

Docker Compose

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    privileged: true
    restart: unless-stopped
    image: ghcr.io/blakeblackshear/frigate:stable-tensorrt
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    shm_size: "2048mb"
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /home/brambo/frigate/config:/config
      - /home/brambo/frigate/config/model_cache:/config/model_cache
      - /mnt/frigate_storage:/media/frigate
      #fix voor blackwell GPU in frigate 0.17, niet meer nodig vanaf 0.18
      - /home/brambo/frigate/config/onnx_fix.py:/opt/frigate/frigate/detectors/plugins/onnx.py
      #added for additional local logging for negative inference time
     # - /home/brambo/frigate/config/base-logging.py:/opt/frigate/frigate/object_detection/base.py
      - type: tmpfs
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "8971:8971"
      - "5000:5000"
      - "8554:8554"
      - "8555:8555"
      - "1984:1984"
    environment:
      FRIGATE_RTSP_PASSWORD: "piep"
      NVIDIA_VISIBLE_DEVICES: "all"
      NVIDIA_DRIVER_CAPABILITIES: "all"
      YOLO_MODELS: "yolov8n-320"
      USE_FP16: "true"

Object Detector

Other

Network connection

Wired

Camera make and model

hikvision

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

---

This is a limitation of the COCO-based models you're using. The COCO labelmap includes dog and horse as separate classes, but ponies visually overlap with both—they're smaller than typical horses and can have proportions similar to large dogs. The model is making a reasonable (but incorrect) guess based on the training data it was exposed to.

Options to reduce these false positives:

1. Use object proportion filters

You can filter out detections that don't match expected dog proportions using min_ratio and max_ratio. These compare the width/height ratio of the bounding box. A pony grazing will typically have a wider, flatter bounding box than a standing dog.(1)

objects:
  filters:
    dog:
      min_ratio: 0.5
      max_ratio: 1.5

(1)

2. Use object area filters

If your ponies consistently produce larger bounding boxes than actual dogs, you can set a max_area for dogs.(1)

objects:
  filters:
    dog:
      max_area: 25000

(1)

3. Create a custom classification model

Frigate supports custom object classification that can assign sub-labels to detected objects. You could train a model to distinguish "pony" from "dog" for objects detected as dog, then filter based on the sub-label.(2)

The workflow involves:

• Creating a model in the Frigate UI under the Classification page
• Selecting dog as the object label to classify
• Defining classes like actual_dog and pony
• Training with example images from your cameras(2)

4. Use zones with object masks

If ponies are only in certain areas, you can use an object mask to prevent dog detections in those regions.(1)

Unfortunately, the base YOLO models (v9, v26, etc.) all use COCO training data, so switching between them won't resolve this fundamental class confusion. Custom classification or filtering by shape/size are your best options.

📚 Sources:

1. Filters
2. Object Classification

---

Have another question? Just tag @inkeep.

[bb61523]

Thanks! any guidance on how to finetune the model myself?

I already have an object classification for our dog. I could create a dog called Pony there as well. But that feel a little like cheating :-)
classification:
custom:
Obi:
enabled: true
name: Obi
threshold: 0.8
object_config:
objects:
- dog
classification_type: sub_label

Is there a possibility to ask the GenAI to verify if this is really a dog? And let genai alter the object if it thinks it's something else?

And are there any good alternatives to the current COCO dataset?

[hawkeye217]

Fine tuning models is out of the scope of the Frigate docs and GitHub support. I'd consult an LLM to walk you through the steps - it's a fairly involved process.

You could use GenAI for object descriptions but GenAI won't change the label that the object detector assigned to it. https://docs.frigate.video/configuration/genai/genai_objects

Open Images (Google) and Objects365 are other popular datasets. You could look on Hugging Face for models based on these datasets, but you may need to convert it to ONNX yourself. Frigate is dataset-agnostic - it just requires the model to be in the correct format with a correct labelmap.

[NickM-27]

The one other option would be to rename the labels to both he animal and have an object classification model set the sub label as dog or pony

[bb61523]

I tried with sub labels dog and pony for a couple days. It works, but it's not a real good solution. The pony's are not always recognized as pony, for example if they're only partially in view or if both are at the same time. So it's working like 50% of the time that I get the notification pony, and false dog the other half.

I might give Open Images (Google) and Objects365 a try. Does anybody already have good results with these?
Maybe off topic, but why not built the option to choose the dataset via de config options in Frigate? :-)

[NickM-27]

but why not built the option to choose the dataset via de config options in Frigate?

Users can run whatever model they want, generally it is only recent times that other datasets are coming in to supported models and many of the supported model architectures still only have freely available models that are COCO.

As a user you already have the option to run whatever model which includes whatever dataset you want