High FFMPEG CPU Usage in Frigate on Raspberry Pi 4 (8GB) – Optimisation Tips?

Hello everyone,

I’m running Home Assistant on a Raspberry Pi 4 (8GB) with an SSD and have recently set up Frigate, which I’m really enjoying so far. I also have a Coral TPU, and I’ve allocated 256MB of video memory in config.txt.

However, Frigate is reporting “Front has high FFMPEG CPU usage” for my only configured camera (“Front”). Before I start adding more cameras, I’d like to optimise my setup and reduce CPU usage where possible.

Here’s what Frigate reports for CPU usage:

And here’s my current Frigate configuration:

mqtt:
  host: core-mosquitto
  user: ********
  password: ********
  topic_prefix: frigate

detectors:
  coral:
    type: edgetpu
    device: usb

ffmpeg:
  hwaccel_args: preset-rpi-64-h264
  output_args:
    record: preset-record-generic-audio-copy

go2rtc:
  streams:
    front:
      - rtsp://192.168.72.151:554/s0

objects:
  # Optional: list of objects to track from labelmap.txt (default: shown below)
  track:
    - person
    - dog
    - car

cameras:
  front:
    detect:
    #  width: 1920
    #  height: 1080
      fps: 5
      enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/front
          roles:
            - detect
            - record
    zones:
      Front_Yard:
        coordinates: 
          0.846,0.352,0.443,0.003,0.254,0.001,0.002,0.235,0.001,0.666,0.157,0.999,0.82,0.999
        loitering_time: 0
      Roadway:
        coordinates: 0.997,0.585,0.995,0.003,0.45,0,0.853,0.356,0.848,0.4
        loitering_time: 1
        inertia: 3
    motion:
      mask: 0.001,0.034,0.175,0.038,0.175,0,0.001,0.001
      threshold: 30
      contour_area: 10
      improve_contrast: true
    review:
      alerts:
        required_zones: Front_Yard
      detections: {}
version: 0.15-1

record:
  enabled: true
  retain:
    days: 14
    mode: all

Does anyone have any tips or best practices for lowering FFMPEG CPU usage on a Raspberry Pi? Any advice would be greatly appreciated!

Thanks in advance!