[Support]: Face identification not accurate

[tbclark3]

Describe the problem you are having

I am trying to get face identification set up in a way that's reliable and useful. I have done the training like you recommended, gradually increasing. There are only two people in the library, and each of us now has about 25 images, half of which were face shots taken from different angles with a phone, and half of which are captured from one of the cameras. I am particularly interested in 3 of the cameras, and I have increased the resolution of the detect stream on those in order to get good face and license plate images. Frigate often gets the identification correct now, but in maybe 10% of the training images it makes egregious mistakes. For example, on a blurry image of the back of my wife's head, it will identify the image as me at 100%. It frequently reports 100% and I'm not sure that's even possible on any image, but on an image of the back of somebody else's head it shouldn't happen. I have played around with the threshold for faces, and haven't been able to improve the problem. I have also played around with the resolution provided, and higher resolution does seem to increase the percentage of correct identifications, but it does not decrease the number of images that are 100% but wrong. I would appreciate some direction.

Version

0.16.0-0b7a33d

What browser(s) are you using?

No response

Frigate config file

auth:
  reset_admin_password: false

face_recognition:
  enabled: true
  model_size: large
  unknown_score: 0.8 #default 0.8, matches at or below this number get marked as unknown
  recognition_threshold: 0.95 #default 0.9, above this number is applied to the person object
  min_faces: 2 #default 1, min face recognitions for the sub label to be applied to the person object
  save_attempts: 250 #default 100, number of images of recognized faces to save for training

mqtt:
  host: mqtt.ferree-clark.org
  port: 1883
  topic_prefix: frigate
  client_id: frigate
  user: frigate
  password: xxxxx

detectors:
  coral0:
    type: edgetpu
    device: usb:0
  coral1:
    type: edgetpu
    device: usb:1

database:
  path: /media/frigate/frigate.db
model:
  path: plus://d692b929ed63f6140955ab7aaf0d248b

tls:
  enabled: false

logger:
  default: info

birdseye:
  enabled: true
  quality: 8
  mode: objects
  restream: true
  inactivity_threshold: 120
  layout:
    max_cameras: 4

genai:
  enabled: true
  provider: gemini
  api_key: xxxx
  model: gemini-2.5-flash
# provider: openai
# api_key: 
#   xxxx
# model: gpt-4o-mini
# prompt: "Analyze the sequence of images containing the {label} from the {camera} camera. If there is more than one {label} say how many there are.  Comment on the appearance of the {label} and consider what they might be doing.  Comment on other objects that appear transient, but do not describe the background since it does not change. Include a 1 or 2 sentence summary of the most important information suitable for a verbal announcement."
  prompt: >
    Prompt:
    You will receive 1 or more sequential images from a camera named {camera}. In
    each case, the camera is mounted on a wall of the house, looking out from the
    house. The images are submitted due to the detection of a {label}, but there may
    be other objects in the images that are just as important. Your task is to:

    1.  Identify,count and describe all human, vehicle, or animal objects. Ignore
    the background. Include the gender and approximate age of any people if you can
    be at least 80% sure of it. If you recognize a logo on a car or person, such as
    Amazon or FedEx, include that in your description and summary.
    2.  Describe what the objects are doing based on their movement across the sequence
    if you can.
    3.  Provide a concise summary announcement suitable for verbal alerts.

    Format:
    Detailed Description: Explain what objects are present and what actions they are
    performing. Mention the {camera} where relevant.
    Summary Announcement: A short, clear sentence or two summarizing the objects present
    and key observations.

    Example Output:
    Detailed Description: A  man enters from the left side of the back driveway, walking
    toward a parked car. They open the passenger door and lean inside.
    Summary Announcement: There is a person in back driveway approaching a parked
    car.

    Ensure accuracy in identifying and interpreting movement while keeping descriptions
    concise and focused only on the relevant activity.



semantic_search:
  enabled: true
  model_size: large

ffmpeg:
  hwaccel_args: preset-nvidia-h264
  output_args:
    record: preset-record-generic-audio-aac

detect:
  enabled: true
  stationary:
    interval: 50
    threshold: 50
    max_frames:
      objects:
        person: 900

motion:
  improve_contrast: false

objects:
  track:
    - person
    - face
    - car
    - bicycle
    - motorcycle
    - license_plate
    - dog
    - cat
    - raccoon
    - rabbit
    - deer
    - bird
    - squirrel
    - amazon
    - usps
    - dhl
    - fedex
    - ups
    - package
  filters:
    person:
      min_area: 1000 #default 0, suggested 5000
      max_area: 200000 #default 100000
      min_score: 0.6 #recommended with frigate+ model 0.65, standard model 0.5
      min_ratio: 0.5 #default 0.5, ratio is width/height
      max_ratio: 2.0 #
      threshold: 0.8 #recommended with frigate+ model 0.85, standard model 0.7
    deer:
      min_score: .6
      threshold: .8
    dog:
      min_score: .7
      threshold: .9
    cat:
      min_score: .7 #recommended with frigate+ model .65
      threshold: .8
    face:
      min_score: .75
    package:
      min_score: .65
      threshold: .9
    license_plate:
      min_score: .6
    amazon:
      min_score: .75
    ups:
      min_score: .75
    fedex:
      min_score: .75
    car:
      min_score: .65
      threshold: .85
    rabbit:
      min_score: 0.5
      threshold: 0.6

record:
  enabled: true
  retain:
    mode: motion
    days: 10
  expire_interval: 360
  alerts:
    retain:
      days: 30
  detections:
    retain:
      days: 30
snapshots:
  enabled: true
  clean_copy: true
  timestamp: false
  bounding_box: true
  crop: false
  retain:
    default: 30
    objects:
      person: 60
      dog: 14
      cat: 14
      deer: 14
      raccoon: 7
      rabbit: 7
      car: 60

live:
  height: 720 #default 720
  quality: 8 #default 8

timestamp_style:
  position: tl
  format: '%m/%d/%Y %H:%M:%S'
  color:
    red: 255
    green: 255
    blue: 255
  thickness: 2

go2rtc:
  webrtc:
    listen: :8555
  streams:
    camera-backyard:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-backyard:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-backyard#audio=opus
    camera-backyard_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-backyard:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-backyard_sub#audio=opus
    camera-street:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-street:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-street#audio=opus
    camera-street_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-street:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-street_sub#audio=opus
    camera-backdoor:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-backdoor:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-backdoor#audio=opus
    camera-backdoor_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-backdoor:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-backdoor_sub#audio=opus
    camera-alley:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-alley:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-alley#audio=opus
#   camera-alley_sub:
#     - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-alley:554/cam/realmonitor?channel=1&subtype=1
#     - ffmpeg:camera-alley_sub#audio=opus
    camera-sidewalk:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-sidewalk:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-sidewalk#audio=opus
    camera-sidewalk_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-sidewalk:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-sidewalk_sub#audio=opus
    camera-frontdoor:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-frontdoor:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-frontdoor#audio=opus
#   camera-frontdoor_sub:
#     - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-frontdoor:554/cam/realmonitor?channel=1&subtype=1
#     - ffmpeg:camera-frontdoor_sub#audio=opus
    camera-uppergarage:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-uppergarage.ferree-clark.org:554/Streaming/Channels/1
    doorbell-front:
      - rtsp://ghanms0001:xxxx@doorbell-front:554/mpeg/media.amp
    doorbell-back:
      - rtsp://ghahic0001:xxxx@doorbell-back:554/mpeg/media.amp
    camera-basement:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-basement:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-basement#audio=opus
    camera-basement_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-basement:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-basement_sub#audio=opus
    camera-lowergarage:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-lowergarage:554/cam/realmonitor?channel=1&subtype=0
      - ffmpeg:camera-lowergarage#audio=opus
    camera-lowergarage_sub:
      - rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@camera-lowergarage:554/cam/realmonitor?channel=1&subtype=1
      - ffmpeg:camera-lowergarage_sub#audio=opus

cameras:
  camera-backyard:
    lpr:
      enabled: False
    objects:
      track:
        - person
#       - car
        - bicycle
        - license_plate
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - amazon
        - usps
        - dhl
        - fedex
        - ups
      filters:
        person:
          min_area: 500 #default 0, suggested 5000
          max_area: 10000 #default 100000
          min_score: 0.6 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.4 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.85 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        amazon:
          min_score: .75
        ups:
          min_score: .75
        fedex:
          min_score: .75
        car:
          min_score: .55
          threshold: .75
        license_plate:
          min_score: .6
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-backyard
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp://localhost:8554/camera-backyard_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
    detect:
      enabled: true
      fps: 6
      width: 704
      height: 480
    record:
      enabled: true
    zones:
      back_driveway:
        coordinates: 385,480,241,388,74,276,41,299,111,385,205,480
      garden:
        coordinates: 551,331,317,213,150,309,414,480
      street:
        coordinates: 0,216,443,33,678,144,645,204,475,97,35,264
    motion:
      mask:
        - 277,335,328,164,198,0,0,0,0,156,26,262,121,372
        - 406,269,541,347,652,192,537,110,421,99,320,146,310,269
        - 559,38,654,44,661,0,565,0
  camera-street:
    lpr:
      enabled: False
    genai:
      objects:
        - person
        - dog
        - cat
        - deer
    objects:
      track:
        - person
        - face
        - car
        - motorcycle
        - bicycle
        - license_plate
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - amazon
        - usps
        - dhl
        - fedex
        - ups
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 200000 #default 100000
          min_score: 0.45 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.65 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .65
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        license_plate:
          min_score: .6
        amazon:
          min_score: .75
        ups:
          min_score: .75
        fedex:
          min_score: .75
        car:
          min_score: .65
          threshold: .85
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-street_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://localhost:8554/camera-street
          input_args: preset-rtsp-restream
          roles:
            - record
    detect:
      enabled: true
      fps: 9
      width: 704
      height: 480
    record:
      enabled: true
    zones:
      back_driveway:
        coordinates: 
          0.969,1,0.879,0.785,0.797,0.772,0.583,0.499,0.483,0.284,0.371,0.252,0.24,0.275,0.106,0.349,0,0.425,0,1
        inertia: 3
        loitering_time: 0
      west_sidewalk:
        coordinates: 
          0,0.41,0.092,0.35,0.163,0.273,0.173,0.189,0.127,0.192,0.092,0.283,0,0.342
        inertia: 3
        loitering_time: 0
        objects:
          - cat
          - deer
          - dog
          - bicycle
          - person
          - rabbit
          - raccoon
          - package
      garden:
        coordinates: 
          0.875,0.71,1,0.935,1,0.487,0.805,0.223,0.547,0.115,0.447,0.125,0.393,0.179,0.446,0.231,0.513,0.281,0.582,0.477,0.809,0.758
        inertia: 3
        loitering_time: 0
    motion:
      mask:
        - 412,233,559,360,704,301,704,110,704,0,597,0,418,98
        - 133,67,335,82,467,0,0,0
  camera-backdoor:
    lpr:
      enabled: False
    objects:
      track:
        - person
        - face
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - amazon
        - usps
        - dhl
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 200000 #default 100000
          min_score: 0.45 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.50 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        amazon:
          min_score: .75
        car:
          min_score: .65
          threshold: .85
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-backdoor_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://localhost:8554/camera-backdoor
          input_args: preset-rtsp-restream
          roles:
            - record
    detect:
      enabled: true
      fps: 8
      width: 704
      height: 480
    record:
      enabled: true
    zones:
      east_sidewalk:
        coordinates: 663,480,451,480,398,148,460,145
    motion:
      mask:
        - 704,418,704,0,515,0
        - 387,239,416,480,0,480,0,0,704,0,294,0
  camera-alley:
    objects:
      track:
        - person
        - face
        - car
        - bicycle
        - motorcycle
        - license_plate
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - amazon
        - usps
        - dhl
        - fedex
        - ups
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 800000 #default 100000
          min_score: 0.5 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.65 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        license_plate:
          min_score: .6
        amazon:
          min_score: .75
        ups:
          min_score: .75
        fedex:
          min_score: .75
        car:
          min_score: .65
          threshold: .85
          max_area: 1000000 
          mask: 0.58,0.076,0.579,0.156,0.683,0.151,0.667,0.073
    ffmpeg:
      inputs:
#       - path: rtsp://localhost:8554/camera-alley_sub
#         input_args: preset-rtsp-restream
#         roles:
#           - detect
        - path: rtsp://localhost:8554/camera-alley
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true
      fps: 8
      width: 1280
      height: 960
    record:
      enabled: true
    zones:
      front_driveway:
        coordinates: 1,1,1,0.85,0.853,0.376,0.209,0.394,0,0.471,0,1
        inertia: 3
        loitering_time: 0
      alley:
        coordinates: 0.924,0.361,1,0.239,0.287,0.275,0.202,0.388
        inertia: 3
        loitering_time: 0
    motion:
      mask:
        - 219,130,704,110,704,0,186,0,0,0,0,210
  camera-sidewalk:
    lpr:
      enabled: False
    objects:
      track:
        - person
        - face
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - amazon
        - usps
        - dhl
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 200000 #default 100000
          min_score: 0.6 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.8 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        amazon:
          min_score: .75
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-sidewalk_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://localhost:8554/camera-sidewalk
          input_args: preset-rtsp-restream
          roles:
            - record
    detect:
      enabled: true
      fps: 8
      width: 704
      height: 480
    record:
      enabled: true
    zones:
      west_sidewalk:
        coordinates: 384,114,426,128,282,480,128,480
    motion:
      mask:
        - 0,480,0,0,381,0,370,126,220,325,107,480
        - 415,0,704,0,704,338,704,283,704,480,401,480
  camera-frontdoor:
    genai:
      objects:
        - person
        - dog
        - cat
        - deer
    objects:
      track:
        - person
        - face
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - squirrel
        - car
        - bicycle
        - motorcycle
        - fedex
        - ups
        - license_plate
        - amazon
        - usps
        - dhl
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 600000 #default 100000
          min_score: 0.55 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.75 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        license_plate:
          min_score: .6
        amazon:
          min_score: .75
        ups:
          min_score: .75
        fedex:
          min_score: .75
        car:
          min_score: .65
          threshold: .85
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-frontdoor
          input_args: preset-rtsp-restream
          roles:
            - record
            - detect
#       - path: rtsp://localhost:8554/camera-frontdoor_sub
#         input_args: preset-rtsp-restream
#         roles:
    detect:
      enabled: true
      fps: 8
      width: 960
      height: 720
    record:
      enabled: true
    zones:
      front_porch:
        coordinates: 0,480,500,480,392,433,394,346,517,248,304,147,0,310
      front_sidewalk:
        coordinates: 0.844,0.147,0.674,0.469,0.51,0.356,0.77,0.108
        inertia: 3
        loitering_time: 0
        objects:
          - bicycle
          - cat
          - deer
          - dog
          - package
          - person
          - rabbit
          - raccoon
          - bird
          - squirrel
    motion:
      mask:
        - 1,0.083,1,0,0,0,0,0.083
        - 0.666,0.769,1,0.856,1,0.344,0.73,0.3
  camera-uppergarage:
    genai:
      enabled: false
    objects:
      track:
        - person
        - face
        - car
        - bicycle
        - license_plate
        - dog
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 200000 #default 100000
          min_score: 0.67 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.85 #recommended with frigate+ model 0.85, standard model 0.7
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        license_plate:
          min_score: .6
        car:
          min_score: .75
          threshold: .9
          max_area: 1000000
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-uppergarage
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
      output_args:
        record: -f segment -segment_time 10 -segment_format mp4 
          -reset_timestamps 1 -strftime 1 -c copy -an
    detect:
      enabled: true
      width: 1024
      height: 768
      fps: 6
    record:
      enabled: true
    motion:
      mask:
        - 409,62,407,50,265,46,263,60
        - 1020,219,1020,0,443,0
    snapshots:
      enabled: true
    zones:
      garage_doorway:
        coordinates: 512,768,0,768,0,451
  camera-lowergarage:
    lpr:
      enabled: False
    genai:
      enabled: false
    objects:
      track:
        - person
        - face
        - dog
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 100000 #default 100000
          min_score: 0.65 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.85 #recommended with frigate+ model 0.85, standard model 0.7
        dog:
          min_score: .7
          threshold: .9
        face:
          min_score: .75
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-lowergarage_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://localhost:8554/camera-lowergarage
          input_args: preset-rtsp-restream
          roles:
            - record
    detect:
      enabled: true
      fps: 8
      width: 704
      height: 480
    record:
      enabled: true
    motion:
      mask: 1,0,1,0.08,0,0.087,0,0
    snapshots:
      enabled: true
    birdseye:
      enabled: false
  camera-basement:
    lpr:
      enabled: False
    genai:
      enabled: false
    objects:
      track:
        - person
        - face
        - dog
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 100000 #default 100000
          min_score: 0.65 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.85 #recommended with frigate+ model 0.85, standard model 0.7
        dog:
          min_score: .7
          threshold: .9
        face:
          min_score: .75
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/camera-basement_sub
          input_args: preset-rtsp-restream
          roles:
            - detect
        - path: rtsp://localhost:8554/camera-basement
          input_args: preset-rtsp-restream
          roles:
            - record
    detect:
      enabled: true
      width: 704
      height: 480
      fps: 8
    record:
      enabled: true
    snapshots:
      enabled: true
    birdseye:
      enabled: false
    motion:
      mask:
        - 704,40,704,0,0,0,0,37
  doorbell-front:
    lpr:
      enabled: False
    genai:
      objects:
        - person
    objects:
      track:
        - person
        - face
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - amazon
        - usps
        - dhl
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 100000 #default 100000
          min_score: 0.65 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.85 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .65 #recommended with frigate+ model .65
          threshold: .7
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        amazon:
          min_score: .75
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/doorbell-front
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true
      width: 1280
      height: 720
      fps: 6
    record:
      enabled: true
    zones:
      front_porch:
        coordinates: 404,698,559,538,674,539,713,720,450,720,0,720,0,671,119,607
    motion:
      mask:
        - 0.121,0.422,0.365,0.387,0.553,0.362,0.664,0,0.002,0.004,0.002,0.376
        - 0.155,0.822,0.367,0.814,0.348,0.628,0.139,0.635
        - 0.865,0.536,0.825,1,1,1,1,0,0.819,0
  doorbell-back:
    lpr:
      enabled: False
    genai:
      objects:
        - person
    objects:
      track:
        - person
        - face
        - dog
        - cat
        - raccoon
        - rabbit
        - deer
        - bird
        - amazon
        - usps
        - dhl
        - package
      filters:
        person:
          min_area: 1000 #default 0, suggested 5000
          max_area: 100000 #default 100000
          min_score: 0.60 #recommended with frigate+ model 0.65, standard model 0.5
          min_ratio: 0.5 #default 0.5, ratio is width/height
          max_ratio: 2.0 #
          threshold: 0.80 #recommended with frigate+ model 0.85, standard model 0.7
        deer:
          min_score: .6
          threshold: .8
        dog:
          min_score: .7
          threshold: .9
        cat:
          min_score: .7 #recommended with frigate+ model .65
          threshold: .8
        face:
          min_score: .75
        package:
          min_score: .65
          threshold: .9
        amazon:
          min_score: .75
    ffmpeg:
      inputs:
        - path: rtsp://localhost:8554/doorbell-back
          input_args: preset-rtsp-restream
          roles:
            - detect
            - record
    detect:
      enabled: true
      width: 1280
      height: 720
      fps: 6
    record:
      enabled: true
    zones:
      back_driveway:
        coordinates: 0.957,1,0,1,0,0.832,0.339,0.665,0.394,0.667,0.493,0.722
        inertia: 3
        loitering_time: 0
      west_sidewalk:
        coordinates: 371,407,284,499,202,487,303,402
    motion:
      mask:
        - 0,0,1280,0,1280,445,574,434,407,431,328,342,0,438
        - 1280,720,1280,433,1010,468,936,616
telemetry:
  network_interfaces:
    - eth0
version: 0.16-0
camera_groups:
  Birdseye:
    order: 1
    icon: LuBird
    cameras: birdseye
classification:
  bird:
    enabled: true
lpr:
  enabled: true
  min_plate_length: 5
  match_distance: 1 #Allow one character variation in matching
  known_plates: #Accepts regex
    Ioniq 5:
      - "CMC5863"

Relevant Frigate log output

There are no errors

Relevant go2rtc log output

There are no errors

FFprobe output from your camera

ffprobe version 7.1.1 Copyright (c) 2007-2025 the FFmpeg developers
  built with gcc 15 (GCC)
  configuration: --prefix=/usr --bindir=/usr/bin --datadir=/usr/share/ffmpeg --docdir=/usr/share/doc/ffmpeg --incdir=/usr/include/ffmpeg --libdir=/usr/lib64 --mandir=/usr/share/man --arch=x86_64 --optflags='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wno-error=incompatible-pointer-types -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' --extra-ldflags='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' --disable-htmlpages --enable-pic --disable-stripping --enable-shared --disable-static --enable-gpl --enable-version3 --enable-libsmbclient --disable-openssl --enable-bzlib --enable-frei0r --enable-chromaprint --enable-gcrypt --enable-gnutls --enable-ladspa --enable-lcms2 --enable-libshaderc --enable-vulkan --disable-cuda-sdk --enable-libaom --enable-libaribb24 --enable-libaribcaption --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libdav1d --enable-libdc1394 --enable-libdvdnav --enable-libdvdread --enable-libfdk-aac --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libharfbuzz --enable-libgme --enable-libgsm --enable-libiec61883 --enable-libilbc --enable-libjack --enable-libjxl --enable-liblc3 --enable-libmodplug --enable-libmp3lame --enable-libmysofa --enable-libopenh264 --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libplacebo --enable-libpulse --enable-libqrencode --enable-librabbitmq --enable-librav1e --enable-librist --enable-librsvg --enable-librubberband --enable-libsnappy --enable-libsvtav1 --enable-libsoxr --enable-libspeex --enable-libssh --enable-libsrt --enable-libtesseract --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvmaf --enable-libvorbis --enable-libv4l2 --enable-libvpx --enable-libwebp --enable-libxml2 --enable-libzimg --enable-libzmq --enable-libzvbi --enable-lto --enable-libvpl --enable-lv2 --enable-vaapi --enable-vdpau --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libvo-amrwbenc --enable-libxvid --enable-openal --enable-opencl --enable-opengl --enable-pthreads --enable-vapoursynth --enable-muxers --enable-demuxers --enable-hwaccels --disable-encoders --disable-decoders --disable-decoder='h264,hevc,libxevd,vc1,vvc' --enable-encoder=',a64multi,a64multi5,aac,libfdk_aac,ac3,adpcm_adx,adpcm_argo,adpcm_g722,adpcm_g726,adpcm_g726le,adpcm_ima_alp,adpcm_ima_amv,adpcm_ima_apm,adpcm_ima_qt,adpcm_ima_ssi,adpcm_ima_wav,adpcm_ima_ws,adpcm_ms,adpcm_swf,adpcm_yamaha,alac,alias_pix,amv,anull,apng,ass,asv1,asv2,av1_amf,av1_nvenc,av1_qsv,av1_vaapi,bitpacked,bmp,cinepak,cljr,dca,dfpwm,dnxhd,dnxhr,dpx,dvbsub,dvdsub,dvvideo,exr,ffv1,ffvhuff,flac,flashsv,flashsv2,flv,g723_1,gif,h261,h263,h263_v4l2m2m,h263p,h264_amf,h264_nvenc,h264_qsv,h264_v4l2m2m,h264_vaapi,hap,hdr,hevc_amf,hevc_nvenc,hevc_qsv,hevc_v4l2m2m,hevc_vaapi,huffyuv,ilbc,jpeg2000,jpegls,libaom,libaom_av1,libcodec2,libgsm,libgsm_ms,libilbc,libjxl,liblc3,libmp3lame,libopencore_amrnb,libopenh264,libopenjpeg,libopus,librav1e,libspeex,libsvtav1,libtheora,libtwolame,libvo_amrwbenc,libvorbis,libvpx_vp8,libvpx_vp9,libwebp,libwebp_anim,libxvid,mjpeg,mjpeg_qsv,mjpeg_vaapi,mlp,mp2,mp2fixed,mpeg1video,mpeg2_qsv,mpeg2_vaapi,mpeg2video,mpeg4,mpeg4_v4l2m2m,msmpeg4v2,msmpeg4v3,msvideo1,nellymoser,opus,pam,pbm,pcm_alaw,pcm_f32be,pcm_f32le,pcm_f64be,pcm_f64le,pcm_mulaw,pcm_s16be,pcm_s16be_planar,pcm_s16le,pcm_s16le_planar,pcm_s24be,pcm_s24le,pcm_s24le_planar,pcm_s32be,pcm_s32le,pcm_s32le_planar,pcm_s8,pcm_s8_planar,pcm_u16be,pcm_u16le,pcm_u24be,pcm_u24le,pcm_u32be,pcm_u32le,pcm_u8,pcx,pgm,pgmyuv,phm,png,ppm,prores,prores_aw,prores_ks,qoi,qtrle,r10k,r210,ra_144,rawvideo,roq,roq_dpcm,rpza,rv10,rv20,s302m,sbc,sgi,smc,snow,sonic,sonic_ls,speedhq,srt,ssa,subrip,sunrast,svq1,targa,text,tiff,truehd,tta,ttml,utvideo,v210,v308,v408,v410,vc1_qsv,vc1_v4l2m2m,vc2,vnull,vorbis,vp8_qsv,vp8_v4l2m2m,vp8_vaapi,vp9_qsv,vp9_vaapi,wavpack,wbmp,webvtt,wmav1,wmav2,wmv1,wmv2,wrapped_avframe,xbm,xface,xsub,xwd,y41p,yuv4,zlib,zmbv,' --enable-decoder=',aac,aasc,libfdk_aac,ac3,acelp_kelvin,adpcm_4xm,adpcm_adx,adpcm_afc,adpcm_agm,adpcm_aica,adpcm_argo,adpcm_ct,adpcm_dtk,adpcm_ea,adpcm_ea_maxis_xa,adpcm_ea_r1,adpcm_ea_r2,adpcm_ea_r3,adpcm_ea_xas,adpcm_g722,adpcm_g726,adpcm_g726le,adpcm_ima_acorn,adpcm_ima_alp,adpcm_ima_amv,adpcm_ima_apc,adpcm_ima_apm,adpcm_ima_cunning,adpcm_ima_dat4,adpcm_ima_dk3,adpcm_ima_dk4,adpcm_ima_ea_eacs,adpcm_ima_ea_sead,adpcm_ima_iss,adpcm_ima_moflex,adpcm_ima_mtf,adpcm_ima_oki,adpcm_ima_qt,adpcm_ima_qt_at,adpcm_ima_rad,adpcm_ima_smjpeg,adpcm_ima_ssi,adpcm_ima_wav,adpcm_ima_ws,adpcm_ms,adpcm_mtaf,adpcm_psx,adpcm_sbpro_2,adpcm_sbpro_3,adpcm_sbpro_4,adpcm_swf,adpcm_thp,adpcm_thp_le,adpcm_vima,adpcm_xa,adpcm_xmd,adpcm_yamaha,adpcm_zork,aic,alac,alias_pix,amrnb,amrwb,amv,anm,ansi,anull,apac,ape,apng,arbc,argo,ass,asv1,asv2,atrac1,atrac3,atrac3al,atrac3p,atrac3pal,aura,aura2,av1,av1_qsv,bethsoftvid,bfi,bink,binkaudio_dct,binkaudio_rdft,bintext,bitpacked,bmp,bmv_audio,bmv_video,bonk,brender_pix,c93,cbd2_dpcm,ccaption,cdgraphics,cdtoons,cdxl,cinepak,clearvideo,cljr,cook,cpia,cscd,cyuv,dca,dds,derf_dpcm,dfa,dfpwm,dirac,dnxhd,dnxhr,dolby_e,dpx,dsd_lsbf,dsd_msbf,dsicinaudio,dsicinvideo,dss_sp,dvaudio,dvbsub,dvdsub,dvvideo,dxa,dxtory,eacmv,eamad,eatgq,eatgv,eatqi,eightbps,eightsvx_exp,eightsvx_fib,escape124,escape130,evrc,exr,ffv1,ffvhuff,ffwavesynth,fits,flac,flashsv,flashsv2,flic,flv,fmvc,fourxm,ftr,g723_1,g729,gdv,gem,gif,gremlin_dpcm,gsm,gsm_ms,gsm_ms_at,h261,h263,h263_v4l2m2m,h263i,h263p,hap,hca,hcom,hdr,hnm4_video,hq_hqa,hqx,huffyuv,hymt,iac,idcin,idf,iff_ilbm,ilbc,imc,indeo2,indeo3,indeo4,indeo5,interplay_acm,interplay_dpcm,interplay_video,ipu,jacosub,jpeg2000,jpegls,jv,kgv1,kmvc,lagarith,libaribb24,libaribcaption,libaom,libaom_av1,libcodec2,libdav1d,libgsm,libgsm_ms,libilbc,libjxl,liblc3,libopencore_amrnb,libopencore_amrwb,libopenh264,libopenjpeg,libopus,librsvg,libschroedinger,libspeex,libvorbis,libvpx_vp8,libvpx_vp9,libzvbi_teletext,loco,lscr,m101,mace3,mace6,mdec,media100,metasound,microdvd,mimic,misc4,mjpeg,mjpeg_qsv,mjpegb,mlp,mmvideo,motionpixels,mp1,mp1float,mp2,mp2float,mp3,mp3adu,mp3adufloat,mp3float,mp3on4,mp3on4float,mpc7,mpc8,mpeg1_v4l2m2m,mpeg1video,mpeg2_qsv,mpeg2_v4l2m2m,mpeg2video,mpeg4,mpeg4_v4l2m2m,mpegvideo,mpl2,msa1,mscc,msmpeg4v1,msmpeg4v2,msmpeg4v3,msnsiren,msp2,msrle,mss1,mss2,msvideo1,mszh,mts2,mv30,mvc1,mvc2,mvdv,mvha,mwsc,mxpeg,nellymoser,nuv,on2avc,opus,paf_audio,paf_video,pam,pbm,pcm_alaw,pcm_bluray,pcm_dvd,pcm_f16le,pcm_f24le,pcm_f32be,pcm_f32le,pcm_f64be,pcm_f64le,pcm_lxf,pcm_mulaw,pcm_s16be,pcm_s16be_planar,pcm_s16le,pcm_s16le_planar,pcm_s24be,pcm_s24daud,pcm_s24le,pcm_s24le_planar,pcm_s32be,pcm_s32le,pcm_s32le_planar,pcm_s64be,pcm_s64le,pcm_s8,pcm_s8_planar,pcm_sga,pcm_u16be,pcm_u16le,pcm_u24be,pcm_u24le,pcm_u32be,pcm_u32le,pcm_u8,pcm_vidc,pcx,pfm,pgm,pgmyuv,pgssub,pgx,phm,photocd,pictor,pjs,png,ppm,prores,prosumer,psd,ptx,qcelp,qdm2,qdmc,qdraw,qoa,qoi,qpeg,qtrle,r10k,r210,ra_144,ra_288,rasc,rawvideo,realtext,rka,rl2,roq,roq_dpcm,rpza,rscc,rv10,rv20,s302m,sami,sanm,sbc,screenpresso,sdx2_dpcm,sgi,sgirle,shorten,simbiosis_imx,sipr,siren,smackaud,smacker,smc,smvjpeg,snow,sol_dpcm,sonic,sp5x,speedhq,speex,srgc,srt,ssa,stl,subrip,subviewer,subviewer1,sunrast,svq1,svq3,tak,targa,targa_y216,tdsc,text,theora,thp,tiertexseqvideo,tiff,tmv,truehd,truemotion1,truemotion2,truemotion2rt,truespeech,tscc,tscc2,tta,twinvq,txd,ulti,utvideo,v210,v210x,v308,v408,v410,vb,vble,vcr1,vmdaudio,vmdvideo,vmnc,vnull,vorbis,vp3,vp4,vp5,vp6,vp6a,vp6f,vp7,vp8,vp8_qsv,vp8_v4l2m2m,vp9,vp9_qsv,vp9_v4l2m2m,vplayer,vqa,vqc,wady_dpcm,wavarc,wavpack,wbmp,wcmv,webp,webvtt,wmav1,wmav2,wmavoice,wmv1,wmv2,wnv1,wrapped_avframe,ws_snd1,xan_dpcm,xan_wc3,xan_wc4,xbin,xbm,xface,xl,xpm,xsub,xwd,y41p,ylc,yop,yuv4,zero12v,zerocodec,zlib,zmbv,'
  libavutil      59. 39.100 / 59. 39.100
  libavcodec     61. 19.101 / 61. 19.101
  libavformat    61.  7.100 / 61.  7.100
  libavdevice    61.  3.100 / 61.  3.100
  libavfilter    10.  4.100 / 10.  4.100
  libswscale      8.  3.100 /  8.  3.100
  libswresample   5.  3.100 /  5.  3.100
  libpostproc    58.  3.100 / 58.  3.100

Frigate stats

{"cameras":{"camera-backyard":{"camera_fps":6.1,"process_fps":6.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":736,"capture_pid":888,"ffmpeg_pid":894,"audio_rms":0.0,"audio_dBFS":0.0},"camera-street":{"camera_fps":9.1,"process_fps":9.1,"skipped_fps":0.0,"detection_fps":9.1,"detection_enabled":true,"pid":745,"capture_pid":900,"ffmpeg_pid":906,"audio_rms":0.0,"audio_dBFS":0.0},"camera-backdoor":{"camera_fps":8.1,"process_fps":8.2,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":752,"capture_pid":912,"ffmpeg_pid":918,"audio_rms":0.0,"audio_dBFS":0.0},"camera-alley":{"camera_fps":8.1,"process_fps":8.0,"skipped_fps":0.0,"detection_fps":1.6,"detection_enabled":true,"pid":763,"capture_pid":924,"ffmpeg_pid":930,"audio_rms":0.0,"audio_dBFS":0.0},"camera-sidewalk":{"camera_fps":8.1,"process_fps":8.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":774,"capture_pid":934,"ffmpeg_pid":940,"audio_rms":0.0,"audio_dBFS":0.0},"camera-frontdoor":{"camera_fps":8.1,"process_fps":8.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":784,"capture_pid":946,"ffmpeg_pid":952,"audio_rms":0.0,"audio_dBFS":0.0},"camera-uppergarage":{"camera_fps":6.0,"process_fps":6.1,"skipped_fps":0.0,"detection_fps":0.1,"detection_enabled":true,"pid":833,"capture_pid":956,"ffmpeg_pid":965,"audio_rms":0.0,"audio_dBFS":0.0},"camera-lowergarage":{"camera_fps":8.1,"process_fps":8.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":844,"capture_pid":969,"ffmpeg_pid":975,"audio_rms":0.0,"audio_dBFS":0.0},"camera-basement":{"camera_fps":8.1,"process_fps":8.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":855,"capture_pid":981,"ffmpeg_pid":987,"audio_rms":0.0,"audio_dBFS":0.0},"doorbell-front":{"camera_fps":6.1,"process_fps":6.1,"skipped_fps":0.0,"detection_fps":0.0,"detection_enabled":true,"pid":866,"capture_pid":994,"ffmpeg_pid":2231806,"audio_rms":0.0,"audio_dBFS":0.0},"doorbell-back":{"camera_fps":6.0,"process_fps":6.0,"skipped_fps":0.0,"detection_fps":6.4,"detection_enabled":true,"pid":877,"capture_pid":1004,"ffmpeg_pid":2170223,"audio_rms":0.0,"audio_dBFS":0.0}},"detectors":{"coral0":{"inference_speed":7.5,"detection_start":0.0,"pid":1135857},"coral1":{"inference_speed":7.41,"detection_start":0.0,"pid":1304722}},"detection_fps":17.2,"embeddings":{"image_embedding_speed":16.57,"image_embedding":0.1,"text_embedding_speed":775.8,"text_embedding":0.1,"face_recognition_speed":28.77,"face_recognition":0.0,"plate_recognition_speed":24.89,"plate_recognition":0.0},"gpu_usages":{"NVIDIA GeForce RTX 4060":{"gpu":"3.0%","mem":"39.44%","enc":"0.0%","dec":"14.0%"}},"cpu_usages":{"frigate.full_system":{"cpu":"18.6","mem":"30.3"},"1":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"/package/admin/s6/command/s6-svscan -d4 -- /run/service"},"17":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6-linux-init-shutdownd"},"18":{"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"},"27":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6rc-oneshot-runner"},"28":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise s6rc-fdholder"},"29":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise certsync"},"30":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise certsync-log"},"31":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise frigate"},"32":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise frigate-log"},"33":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc"},"34":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc-healthcheck"},"35":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise go2rtc-log"},"36":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise nginx"},"37":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-supervise nginx-log"},"48":{"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"},"49":{"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 ../.. --"},"100":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/go2rtc"},"101":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/nginx"},"103":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/frigate"},"104":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"s6-log -b -- T 1 n0 s10000000 T /dev/shm/logs/certsync"},"116":{"cpu":"35.6","cpu_average":"34","mem":"0.5","cmdline":"/usr/local/go2rtc/bin/go2rtc -config=/dev/shm/go2rtc.yaml"},"135":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"bash ./run.user go2rtc-healthcheck"},"136":{"cpu":"11.5","cpu_average":"9","mem":"8.1","cmdline":"python3 -u -m frigate"},"144":{"cpu":"0.0","cpu_average":"0","mem":"0.1","cmdline":"nginx: master process nginx"},"374":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"375":{"cpu":"0.0","cpu_average":"0","mem":"0.3","cmdline":"nginx: worker process"},"376":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: worker process"},"377":{"cpu":"0.0","cpu_average":"0","mem":"0.1","cmdline":"nginx: worker process"},"378":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"nginx: cache manager process"},"508":{"cpu":"0.0","cpu_average":"0","mem":"0.6","cmdline":"python3 -u -m frigate"},"517":{"cpu":"11.0","cpu_average":"9","mem":"1.4","cmdline":"frigate.recording_manager"},"526":{"cpu":"1.9","cpu_average":"1","mem":"1.1","cmdline":"frigate.review_segment_manager"},"551":{"cpu":"73.2","cpu_average":"4","mem":"17.9","cmdline":"frigate.embeddings_manager"},"577":{"cpu":"23.5","cpu_average":"8","mem":"1.2","cmdline":"frigate.output       "},"587":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"594":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"601":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"608":{"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 1280x960 -i pipe: -threads 1 -f mpegts -s 960x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"615":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"622":{"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 960x720 -i pipe: -threads 1 -f mpegts -s 960x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"629":{"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 1024x768 -i pipe: -threads 1 -f mpegts -s 960x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"636":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"643":{"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 704x480 -i pipe: -threads 1 -f mpegts -s 1056x720 -codec:v mpeg1video -q 8 -bf 0 pipe:"},"650":{"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:"},"657":{"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:"},"664":{"cpu":"7.0","cpu_average":"1","mem":"0.2","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:"},"675":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"bash ./run.user certsync"},"736":{"cpu":"0.9","cpu_average":"1","mem":"1.4","cmdline":"frigate.process:camera-backyard"},"745":{"cpu":"20.9","cpu_average":"2","mem":"1.4","cmdline":"frigate.process:camera-street"},"752":{"cpu":"1.3","cpu_average":"1","mem":"1.4","cmdline":"frigate.process:camera-backdoor"},"763":{"cpu":"4.2","cpu_average":"3","mem":"1.5","cmdline":"frigate.process:camera-alley"},"774":{"cpu":"1.3","cpu_average":"1","mem":"1.3","cmdline":"frigate.process:camera-sidewalk"},"784":{"cpu":"2.1","cpu_average":"2","mem":"1.5","cmdline":"frigate.process:camera-frontdoor"},"833":{"cpu":"1.9","cpu_average":"1","mem":"1.4","cmdline":"frigate.process:camera-uppergarage"},"844":{"cpu":"1.1","cpu_average":"1","mem":"1.3","cmdline":"frigate.process:camera-lowergarage"},"855":{"cpu":"1.2","cpu_average":"1","mem":"1.4","cmdline":"frigate.process:camera-basement"},"866":{"cpu":"1.1","cpu_average":"1","mem":"1.5","cmdline":"frigate.process:doorbell-front"},"877":{"cpu":"19.9","cpu_average":"1","mem":"1.4","cmdline":"frigate.process:doorbell-back"},"888":{"cpu":"1.3","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-backyard"},"894":{"cpu":"2.6","cpu_average":"2","mem":"0.8","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-backyard_sub -r 6 -vf fps=6,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"899":{"cpu":"3.0","cpu_average":"3","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-backyard -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-backyard@%Y%m%d%H%M%S%z.mp4"},"900":{"cpu":"1.8","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-street"},"906":{"cpu":"3.4","cpu_average":"3","mem":"1.8","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-street_sub -r 9 -vf fps=9,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"911":{"cpu":"3.4","cpu_average":"3","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-street -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-street@%Y%m%d%H%M%S%z.mp4"},"912":{"cpu":"1.5","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-backdoor"},"918":{"cpu":"2.9","cpu_average":"3","mem":"0.8","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-backdoor_sub -r 8 -vf fps=8,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"923":{"cpu":"3.2","cpu_average":"3","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-backdoor -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-backdoor@%Y%m%d%H%M%S%z.mp4"},"924":{"cpu":"4.4","cpu_average":"4","mem":"1.3","cmdline":"frigate.capture:camera-alley"},"930":{"cpu":"9.6","cpu_average":"10","mem":"1.3","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-alley -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-alley@%Y%m%d%H%M%S%z.mp4 -r 8 -vf fps=8,scale_cuda=w=1280:h=960,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"934":{"cpu":"1.5","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-sidewalk"},"940":{"cpu":"3.0","cpu_average":"3","mem":"0.9","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-sidewalk_sub -r 8 -vf fps=8,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"945":{"cpu":"3.3","cpu_average":"3","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-sidewalk -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-sidewalk@%Y%m%d%H%M%S%z.mp4"},"946":{"cpu":"2.5","cpu_average":"2","mem":"1.3","cmdline":"frigate.capture:camera-frontdoor"},"952":{"cpu":"8.2","cpu_average":"8","mem":"2.1","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-frontdoor -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-frontdoor@%Y%m%d%H%M%S%z.mp4 -r 8 -vf fps=8,scale_cuda=w=960:h=720,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"956":{"cpu":"2.3","cpu_average":"2","mem":"1.3","cmdline":"frigate.capture:camera-uppergarage"},"965":{"cpu":"4.6","cpu_average":"4","mem":"1.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-uppergarage -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c copy -an /tmp/cache/camera-uppergarage@%Y%m%d%H%M%S%z.mp4 -r 6 -vf fps=6,scale_cuda=w=1024:h=768,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"969":{"cpu":"1.5","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-lowergarage"},"975":{"cpu":"3.0","cpu_average":"3","mem":"0.8","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-lowergarage_sub -r 8 -vf fps=8,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"980":{"cpu":"2.3","cpu_average":"2","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-lowergarage -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-lowergarage@%Y%m%d%H%M%S%z.mp4"},"981":{"cpu":"1.5","cpu_average":"1","mem":"1.3","cmdline":"frigate.capture:camera-basement"},"987":{"cpu":"3.1","cpu_average":"3","mem":"1.6","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-basement_sub -r 8 -vf fps=8,scale_cuda=w=704:h=480,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"992":{"cpu":"2.5","cpu_average":"2","mem":"0.2","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/camera-basement -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/camera-basement@%Y%m%d%H%M%S%z.mp4"},"994":{"cpu":"2.2","cpu_average":"2","mem":"1.3","cmdline":"frigate.capture:doorbell-front"},"1004":{"cpu":"2.4","cpu_average":"2","mem":"1.3","cmdline":"frigate.capture:doorbell-back"},"1135857":{"cpu":"4.8","cpu_average":"0","mem":"4.9","cmdline":"frigate.detector.coral0"},"1304722":{"cpu":"4.8","cpu_average":"0","mem":"5.7","cmdline":"frigate.detector.coral1"},"2170223":{"cpu":"4.3","cpu_average":"4","mem":"0.9","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/doorbell-back -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/doorbell-back@%Y%m%d%H%M%S%z.mp4 -r 6 -vf fps=6,scale_cuda=w=1280:h=720,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"2231806":{"cpu":"4.0","cpu_average":"4","mem":"1.0","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -loglevel warning -threads 2 -hwaccel cuda -hwaccel_output_format cuda -user_agent FFmpeg Frigate/0.16.0-0b7a33d -rtsp_transport tcp -timeout 10000000 -i rtsp://localhost:8554/doorbell-front -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac /tmp/cache/doorbell-front@%Y%m%d%H%M%S%z.mp4 -r 6 -vf fps=6,scale_cuda=w=1280:h=720,hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5 -threads 2 -f rawvideo -pix_fmt yuv420p pipe:"},"2420224":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"sleep 9999"},"2436854":{"cpu":"3.7","cpu_average":"3","mem":"0.9","cmdline":"/usr/lib/ffmpeg/7.0/bin/ffmpeg -hide_banner -f rawvideo -pix_fmt yuv420p -video_size 1280x720 -r 10 -i /tmp/cache/birdseye -c:v h264_nvenc -g 50 -profile:v high -level:v auto -preset:v p2 -tune:v ll -rtsp_transport tcp -f rtsp rtsp://127.0.0.1:8554/2531bc4fe13ccf35304fb0b8609bf29b"},"2441766":{"cpu":"0.0","cpu_average":"0","mem":"0.0","cmdline":"sleep 30s"}},"service":{"uptime":171316,"version":"0.16.0-0b7a33d","latest_version":"0.15.2","storage":{"/media/frigate/recordings":{"total":7630895.7,"used":1902190.9,"free":5727428.3,"mount_type":"btrfs"},"/media/frigate/clips":{"total":7630895.7,"used":1902190.9,"free":5727428.3,"mount_type":"btrfs"},"/tmp/cache":{"total":1430.5,"used":127.9,"free":1302.6,"mount_type":"tmpfs"},"/dev/shm":{"total":512.0,"used":358.8,"free":153.2,"mount_type":"tmpfs"}},"temperatures":{},"last_updated":1753905082},"processes":{"recording":{"pid":517},"review_segment":{"pid":526},"go2rtc":{"pid":116},"embeddings":{"pid":551}}}

Install method

Docker Compose

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    privileged: false
    restart: "unless-stopped"
    image: ghcr.io/blakeblackshear/frigate:0.16.0-rc1-tensorrt
#   image: ghcr.io/blakeblackshear/frigate:0.15.2-tensorrt
    shm_size: "512mb"
    devices:
      - /dev/bus/usb/006/:/dev/bus/usb/006/
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /docker/frigate/config:/config
      - /export/frigate:/media/frigate
      - type: tmpfs 
        target: /tmp/cache
        tmpfs:
          size: 1500000000
    environment:
      FRIGATE_RTSP_PASSWORD: "xxxx"
      PLUS_API_KEY: "xxxx"
    networks:
      br0-network_br0-network:
        ipv4_address:  10.1.10.39
      vlan4-network_vlan4-network:
        ipv4_address: 10.1.4.39
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
        limits:
          memory: 16G
          cpus: '12.0'
  stunnel:
    container_name: frigate_stunnel
    privileged: false
    depends_on: 
      - frigate
    deploy:
      resources:
        limits:
          memory: 1G
          cpus: "1.0"
    image: dweomer/stunnel
    network_mode: service:frigate
    environment:
      STUNNEL_SERVICE: "frigate"
      STUNNEL_ACCEPT: "443"
      STUNNEL_CONNECT: "8971"
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /etc/letsencrypt/live/ferree-clark.org/fullchain.pem:/etc/stunnel/stunnel.pem:ro
      - /etc/letsencrypt/live/ferree-clark.org/privkey.pem:/etc/stunnel/stunnel.key:ro
networks:
  br0-network_br0-network:
    external: true
  vlan4-network_vlan4-network:
    external: true

Object Detector

Coral

Network connection

Wired

Camera make and model

Amcrest various models

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[NickM-27]

There are only two people in the library, and each of us now has about 25 images, half of which were face shots taken from different angles with a phone, and half of which are captured from one of the cameras

This is incorrect. It is only recommended to have at most 5 phone images for each person.

When it comes to detecting the back of the head as a face, you may need to increase the detection_threshold for faces to reduce those false positives.

Once a face is detected, it is used to generate an embedding that is compared to the trained image embedding. 100% match means there may be some other blurry or otherwise difficult images that lead to the strong match.

It seems to me like the first recommended step is revisiting the training data

[kirashogun]

I have similar issues, seems like a crap shoot. It would be extremely confident in making ridiculous mistakes mistaken my face for my wifes (like 99% confident) and even worse, total strangers faces for mine with 90%+ confidence. Also the images are extremely blurry for some reason even I'm using the intercom camera on full HD resolution.

In the end i went back to Double Take with Compreface. Seems much more accurate and makes more sense than the current frigate implementation. Hopefully we can see improvements from here onwards.

[kirashogun]

That is strange. Remember the only variable for me is frigate 0.16 vs compreface. Camera is not a variable. It's a 1080p h264 rtsp stream that's high quality. The face captured by double take is high res. The clips captured by frigate is high res. I'm really stumped. 🤔

I am using a custom statically compiled ffmpeg as the one that came with 0.15 did not support lunar lake gpu decode. But it did not affect double take nor the video captured by frigate.

[NickM-27]

I don't know what compreface does behind the scenes, perhaps it has some upscaling or something that helps in this case. I know the large model in frigate is a resnet100-arcface model which is similar to compreface's arcface model.

All I know is that the image shown above is quite unclear, I'm not sure I would recognize someone myself if the image looked like that and I didn't have the context of when something happened.

If compreface works for you, I don't really see a reason you'd need to change over, I am just trying to make it clear that they're not 1-1, you can't just use the same training images for both and expect similar results.

[kirashogun]

The snapshots captured by frigate itself is high res. It's only facial capture module that's butchering the image quality on the snapshots used for facial recognition on my setup for some reason.

There's a huge difference between snapshot images in the tracked objects tab vs images in the facial recognition tab.

To me there is a bug in the face module, it's not the camera, it's not double take. There is no image enhancement, the camera is not compressing more all of a sudden. You get my drift.

I like to use frigate if I can as it is quite a bit faster.

[NickM-27]

There is no image enhancement

this is objectively false. there are multiple steps required in adjusting the image before it can be used with the face recognition models, so they likely are doing some things differently regardless of if that affects this specific case or not.

To me there is a bug in the face module

I am not sure how there could be a bug that wouldn't affect everyone, especially in my case where I am using a considerably smaller detect stream than you.

Also it is worth mentioning that the image that is actually sent to the face recognizer has absolutely no compression as it is just a direct crop from the detect stream (which is exactly what is used for the snapshot). The only compression that happens, ever, would be the compression to webp for the images that are shown in the frontend.

The default compression level for webp in cv2 is "visually lossless" meaning it aims to have no change in visual quality, and it uses a standard that would not change based on the size of the image. So it would not change for my smaller images vs your larger ones.

[kyle4269]

It seems like in 0.16, under tracked object details, it shows an image that's fixed at 175 x 175 resolution. Maybe it's just a UI thing? The faces it extracts from these shots some are 180 x 220 pixels, sometimes they are 79 x 80 pixels.

That image is entirely unrelated to the images that are used for face recognition.

Those images are quite blurry especially for that resolution, makes it seem like the camera has too much compression. You could maybe try increasing the camera bandwidth or reducing motion smoothing if that's an option

For example, here's my camera at 1280x720 stream, much smaller face that is considerably more clear

I was struggling with blurry images also and increasing the bandwidth(bitrate) of the camera was the issue.

[tbclark3]

I still have a problem with inaccurate detection and, like @kirashogun, inexplicably blurry images. I was intrigued by the bitrate suggestion, but my streams are configured for 8192, which should be more than enough. I am now wondering if my fps could be misconfigured or suboptimally configured. I am using the camera's mains tream for roles record and detect, configured at 2592*1944, CBR, 20 FPS, bitrate 8192, frame interval 40. This snippet is from my config, which should provide a reasonable image for detection. Maybe I am missing something?

ffmpeg:
  inputs:
    - path: rtsp://localhost:8554/camera-frontdoor
      input_args: preset-rtsp-restream
      roles:
        - record
        - detect
detect:
  enabled: true
  fps: 8
  width: 960
  height: 720
record:
  enabled: true

[NickM-27]

To be clear, I haven't yet seen anything I would consider inexplicably blurry image. Realistically, we would need to see a snapshot, that snapshot cropped down to the face, and then the face that Frigate shows with noticeable difference in blurriness for that to make sense. This is of course difficult for an end user to do because the snapshot and face images are not taken from the exact same frame, so it would be guessing at best.

Right now it just seems like general motion that causes the blur, potentially due to bitrate or cameras motion interpolation for compression. In your case, an i-frame of twice the frame rate could cause blurriness, this means that a fresh frame is only sent every 2 seconds. In between those two seconds it uses motion data to get the frame. I personally am running the detect stream at 1 second i-frame interval.

Depending on your camera you can also look at other things like if the shutter speed is inappropriate, though in the daytime I just leave mine on auto.

[NickM-27]

I still have a problem with inaccurate detection

also, to be clear, is it your detection that is inaccurate or recognition? these are distinctly different things, and it is important to be clear about which step is not working correctly

[tbclark3]

Sorry, I meant recognition. Detection works fine.

[shinyzard123]

I had to move from my camera's D1 sub stream to the second sub stream (720p) and with 5 fps / 5 interval face recognition is working really well, differentiating between adults and even kids that look fairly similar.

The UI definitely takes some getting used to but once you have a flow it makes a lot of sense and I am happy with how things are performing.