Frigate not detecting person events (Tapo C200 and C220 camera's)

Configuration
1

Hi,
I’m using Frigate in combination with my Tapo C2xx camera’s but the person detection isn’t working correctly.

Every now and then I get an event detected, but most of the time it’s not triggering any events/recordings.

Anything I’m missing here (only added the C200 at the moment)

mqtt:
  host: 192.168.0.101
  port: 1883
  user: Kab00mPwd-56
  password: hberznjry,ku,78949zhrnrnzrnr

cameras:
  Tapo1:
    ffmpeg:
      inputs:
        - path: rtsp://camviewer:[email protected]:554/stream2
          roles:
            - rtmp
            - detect
            - record
    rtmp:
      enabled: true
    detect:
      width: 640
      height: 360
      fps: 15
    record:
      enabled: true
      retain:
        days: 15
        mode: active_objects
      events:
        pre_capture: 10
        post_capture: 10
        objects:
          - person
          - dog
          - cat
    snapshots:
      enabled: true
      timestamp: true
      retain:
        default: 10
        objects:
          person: 15

birdseye:
  enabled: false
2

You should define a detector within frigate.yaml.
If you don’t have a TPU but let the CPU do the detection add like:

detectors:
  cpu1:
    type: cpu
    num_threads: 3

Anyway, using the CPU for object detection will put some extra load on your CPU and is not recommended. While this may be still ok with only one camera adding more cameras will sooner or later bring your system down.

If you don’t own a TPU at least consider to configure the OpenVINO Detector.

It is also advised to enable hardware acceleration on the GPU. While this will not help with object detection it will help to take some load off the CPU when decompressing video streams.

  • More information on supported detectors HERE.
  • Additional recommended readings on GPU hardware acceleration, detectors etc. HERE.
3

thanks!!! I inserted the detectors part, so i’ll leave that running for a while and see if it gets better.
I indeed do not have a TPU, and I’m running on a virtual machine within a Win11 on a celeron NUC… So I guess that’s not ideal either for this sort of use

4

That is indeed far from ideal.

Just monitor the system load for a while and see whether it will work out for you.

1 Like
5

thanks! Appreciate the help! have a nice weekend

1 Like
6

just to clarify, CPU detector is defined by default if one is not defined, so this won’t be the reason that detection isn’t working. If I had to guess, like you suggested, detection is not working well simply because it is running on CPU and this is too slow to work correctly. OpenVino in CPU mode should work better here

1 Like