[Support]: Packages not detected

[USER19464727]

Describe the problem you are having

I have Frigate working well and I am now using Frigate+. It is successfully detecting people, cars, FedEx, license plates, etc. I even have Double-Take and codeproject.ai involved and updating the label for people to assign a name to them in Frigate. However, the one thing it won't do is detect a package. For the life of me I'm not sure what is wrong. I have "package" as part of the objects to track. It will put a motion box around it but never tracks it as an object. I've tried adjusting the min_area, min_ratio, max_ratio, min_score, threshold. I've tried putting the box in different positions in frame. It's a large-enough package that it should be recognized (see image).

Version

0.14.1

What browser(s) are you using?

Brave 1.73.89

Frigate config file

mqtt:
  host: core-mosquitto
  user: (REDACTED)
  password: (REDACTED)

database:
  path: /config/frigate.db

detectors:
  coral:
    type: edgetpu
    device: usb

model:
  path: plus://(REDACTED)

audio:
  enabled: true
  max_not_heard: 30
  min_volume: 700
  listen:
    - speech
    - scream
    - yell
    - crying
    - bark
    - fire_alarm

birdseye:
  enabled: true
  restream: false
  mode: objects
  quality: 31

ffmpeg:
  input_args: preset-rtsp-restream-low-latency
  output_args:
    record: preset-record-generic-audio-aac

motion:
  enabled: true

detect:
  enabled: true
  fps: 5
  stationary:
    interval: 50
    threshold: 50

objects:
  track:
    # People
    - person
    - face
    # Vehicles
    - car
    - bicycle
    - motorcycle
    - license_plate
    # Animals
    - deer
    - dog
    - cat
    # Deliveries
    - amazon
    - fedex
    - usps
    - ups
    - package
  filters:
    dog:
      min_score: 0.7
      threshold: 0.9
    cat:
      min_score: 0.65
      threshold: 0.8
    face:
      min_score: 0.7
    package:
      min_area: 0
      min_ratio: 0
      max_ratio: 1000
      min_score: 0
      threshold: 0.2
    license_plate:
      min_score: 0.6
    amazon:
      min_score: 0.75
    ups:
      min_score: 0.75
    fedex:
      min_score: 0.75
    person:
      min_score: 0.65
      threshold: 0.85
    car:
      min_score: 0.65
      threshold: 0.90

review:
  alerts:
    labels:
      - person
      - face
      - car
      - bicycle
      - motorcycle
      - license_plate
      - deer
      - amazon
      - fedex
      - usps
      - ups
      - package
      - crying # audio
      - yell # audio
      - scream # audio
  detections:
    labels:
      - bark # audio
      - fire_alarm # audio
      - speech # audio

go2rtc:
  streams:
    doorbell_latest: #for double-take full-res snapshots
      - onvif://(REDACTED):(REDACTED)@192.168.50.219:8000?subtype=MediaProfile00000&snapshot
    doorbell_main:
      - rtsp://(REDACTED):(REDACTED)@192.168.50.219:554/Preview_01_main#backchannel=0
      - rtsp://(REDACTED):(REDACTED)@192.168.50.219:554/Preview_01_sub
      - ffmpeg:doorbell_main#audio=opus#hardware
    doorbell_sub:
      - rtsp://(REDACTED):(REDACTED)@192.168.50.219:554/Preview_01_sub
      - ffmpeg:doorbell_sub#audio=opus#hardware

record:
  enabled: true
  retain:
    days: 7
    mode: motion

snapshots:
  enabled: true
  bounding_box: true
  crop: true
  height: 200
  retain:
    default: 30
  quality: 100

cameras:
  doorbell:
    enabled: true
    ui:
      order: 0
    audio:
      enabled: true
      min_volume: 500
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/doorbell_main
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/doorbell_sub
          roles:
            - detect
            - audio
    objects:
      mask: 
        0.279,0.372,0.481,0.439,0.48,0.611,0.399,0.615,0.412,0.772,0.296,0.776,0.268,0.57
    review:
      alerts:
        labels:
          - person
          - dog
          - cat
          - package
        required_zones:
          - entryway
          - top-of-steps
      detections:
        labels: []
    motion:
      mask: 0,0.051,1,0.051,1,0,0,0
    zones:
      top-of-steps:
        coordinates: 0,0.915,0.163,0.747,0.861,0.678,1,0.693,1,1,0,1
        loitering_time: 0
        inertia: 3
      entryway:
        coordinates: 0.837,0.055,1,0.261,1,1,0.177,1,0.103,0.483,0.23,0.057
        loitering_time: 0
        inertia: 3

Relevant Frigate log output

N/A

Relevant go2rtc log output

N/A

FFprobe output from your camera

N/A

Frigate stats

N/A

Install method

HassOS Addon

docker-compose file or Docker CLI command

N/A

Object Detector

Coral

Network connection

Doorbell camera is WiFi, everything else wired

Camera make and model

Reolink WiFi Doorbell

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[NickM-27]

How many images have you uploaded with package annotated? In my experience brown boxes work best, but after annotating it does get better

[USER19464727]

Only a few so far, but I thought it should still detect some packages regardless of my individual trainings ("This model will support all label types available even if you do not submit any examples for those labels."). It detected FedEx and I hadn't trained any images for that yet.

[NickM-27]

In general but it can depend on the package, camera angle, etc like I said