Cats being detected as persons

[ronrath]

Person detection is working great and the system consistently detects the cats, however, it detects the cats as people.

Version 2.3 Running in supervised install of HA on Pi with Raspian OS
Camera is Amcrest IP8M-T2599EW

Frigate config file:

detectors:
  coral:
    type: edgetpu
    device: usb

cameras:
  garage:
    ffmpeg:
      inputs:
        - path: rtsp://<UID:PWD@<IP address:port>/cam/realmonitor?channel=1&subtype=1
          roles:
            - detect
            - rtmp
        - path: rtsp://<UID:PWD@<IP address:port>/live
          roles:
            - record
    detect:
      width: 704
      height: 480
      fps: 5
    objects:
      track:
        - person
        - cat
      filters:
        person:
          threshold: 0.5
    motion:
      mask:
        - 700,0,700,46,515,46,515,0
    record:
      enabled: True
      retain_days: 0
      events:
        retain:
          default: 10
    snapshots:
      enabled: True
      timestamp: False
      bounding_box: True
      retain:
        default: 5

birdseye:
  enabled: True
  width: 1280
  height: 720
  # Optional: Encoding quality of the mpeg1 feed (default: shown below)
  # 1 is the highest quality, and 31 is the lowest. Lower quality feeds utilize less CPU resources.
  quality: 3
  # Optional: Mode of the view. Available options are: objects, motion, and continuous
  #   objects - cameras are included if they have had a tracked object within the last 30 seconds
  #   motion - cameras are included if motion was detected in the last 30 seconds
  #   continuous - all cameras are included always
  mode: motion

[blakeblackshear]

This is probably because the threshold for person has been lowered to 0.5. It's just going to happen sometimes if you are saying you only want to be 50% sure it's a person.

[ronrath]

I'm not sure what changed but late yesterday it started working correctly. The 50% could be the problem, each of the wrong detections (cat detected as person) were between 55-65% and the now correct detections are all 75-85%. I'll bump up the threshold a bit on the person detection and see if that helps further. Thanks!