Frigate taking up a lot of CPU, but it's probably my config

[gatorreina]

I have seven cameras and one Coral ai. My frigate.process (s) jump as high as 75% individually and when someone is front of a camera my uswgi can go to 100%. I am using an almost identical config for all of my cameras as below. Is there something I am doing wrong that could be driving my high CPU usage?

mqtt:
host: minisforum1
port: 1883
topic_prefix: frigate
cameras:
cochera1:
ffmpeg:
hwaccel_args:
- -hwaccel
- vaapi
- -hwaccel_device
- /dev/dri/renderD128
- -hwaccel_output_format
- yuv420p
output_args:
record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac
inputs:
- path: rtsp://admin:pass@192.168.50.11:554/cam/realmonitor?channel=1&subtype=1 # Amcr IP5M-T1179EW subs
roles:
- detect
- rtmp
- path: rtsp://admin:pass@192.168.50.11:554/cam/realmonitor?channel=1&subtype=0 # Amcr IP5M-T1179EW subs
roles:
- record
detect:
width: 704
height: 480
fps: 5
objects:
track:
- person
snapshots:
enabled: True
retain:
default: 4
record:
enabled: True
events:
retain:
default: 5
pre_capture: 3
zones:
zone_0:
coordinates: 0,118,42,122,53,234,99,260,276,203,383,190,673,311,896,464,896,672,0,672 #0,0,896,0,896,672,0,672
motion:
mask:
- 704,0,704,50,511,47,509,0

I am also using a coral AI USB accelerator:

detectors:
coral:
type: edgetpu

[NickM-27]

What CPU are you using and have you verified hwaccel is working? (Wouldn't explain the CPU jump, but could explain generally high usage)

What inference speed are you seeing in the frigate debug screen?

[gatorreina]

Beelink 8th Generation Intel i5-8279U Processor（up to 4.1GHz）,Beelink SEi8 with 8G DDR4 RAM/256GB M.2 NVME SSD,Supports 6MB Smart Cache,4K Dual HDMI,WIFI5,BT5.0

How do I verify if hwaccel is working?

Inference speed: 11.93

[NickM-27]

intel_gpu_top in the host or container should work