Q: CPU load does not trottle back after motion?

Describe the bug CPU load does not throttle back after motion(?)

The load on my CPU after restart of the docker container stays around 2.xx and after detection on one or multiple cameras the load never drops below 3.xx. It does not throttle down(?)

Am I alone with this issue?

Version of frigate 0.7.1-amd64

Config file

web_port: 5000

tensorflow_device: usb

mqtt:
  host: 192.168.1.10
  user: username
  password: password
  topic_prefix: frigate

save_clips:
  max_seconds: 300

objects:
  track:
    - person
  filters:
    person:
      min_area: 25000
      max_area: 100000
      threshold: 0.8
      min_score: 0.5

cameras:
  cam9_uppfart:
    ffmpeg:
      input: rtsp://usr:passwd@192.168.1.38:554/Streaming/Channels/1
      hwaccel_args:
        - -hwaccel
        - vaapi
        - -hwaccel_device
        - /dev/dri/renderD128
        - -hwaccel_output_format
        - yuv420p
    take_frame: 1
    fps: 8
    best_image_timeout: 60
    save_clips:
      enabled: False
      pre_capture: 30
    snapshots:
      show_timestamp: True
      draw_zones: True
      draw_bounding_boxes: True

Computer Hardware

• OS: Debian
• Virtualization: No, bare metal
• Coral Version: USB
• Network Setup: Wired

Camera Info:

• Manufacturer: Hikvision
• Model: DS-2CD2T42WD-I5
• Resolution: 2688x1520
• FPS: 8

[blakeblackshear]

Yes. Masks will help with that. In the future, the motion detection will be configurable too.

[tubalainen]

I made polygon masks for all my cameras and it seems to have made some difference! Thanks for the tip. I will make a PR with updated info on how to make a mask using the tool described in the example config. There´s still quite abstract for some users (including my self) how to get it going 😃.