[Config Support]: Two cameras but getting message "Drive has high FFMPEG CPU usage (22%)" #13259
-
Describe the problem you are havingI've looked through a lot of the recommended config settings to try and solve this but I appreciate I may have made an error. I only have two cameras: a 4k camera on the drive (RLC-820A) and one on my front lawn (RLC-520A) However, I'm constantly at about 18 - 22% CPU usage. I've read some of the guides and understand that due to detections etc CPU usage may be higher than expected even with Coral support but I've also read some updates from people with triple the number of cameras pulling the same figures, so I wondered if there was some simple mistake I've made. I've matched the pixel height and width of the camera on the config, as well as putting them both on 5 fps. I've tried disabling object detection on the "front" camera to decrease demand (again, possibly a misconfiguration on my part). Any assistance gratefully received. Version0.14.0-da913d8 Frigate config filemqtt:
enabled: false
detectors:
coral:
type: edgetpu
device: pci
cameras:
Drive: # <--- name assigned
enabled: true
ffmpeg:
hwaccel_args: preset-vaapi # < --- leave this on as there will always be CPU/GPU use by frigate
inputs:
# high res stream
- path: rtsp://admin:[email protected]/Preview_01_main
roles:
- record
- audio # < --- this allows audio on exports/recordings
# low res stream
- path: rtsp://admin:[email protected]/Preview_01_sub
roles:
- detect
output_args:
record: preset-record-generic-audio-copy
detect:
width: 3840
height: 2160
fps: 5
objects:
track:
- person
- cat
mask:
- 0.068,0.579,0.163,0.511,0.221,0.457,0.291,0.393,0.262,0.347,0.2,0.306,0.148,0.308,0.09,0.352,0.036,0.41,0.024,0.498
- 0.156,0.003,0.406,0.007,0.407,0.141,0.192,0.13,0.159,0.124
- 0.452,0.006,0.495,0.008,0.488,0.231,0.445,0.231
- 0.75,0.03,0.847,0.066,0.834,0.209,0.745,0.185
- 0.589,0.004,0.587,0.134,0.51,0.121,0.513,0.002
- 0.094,0.474,0.324,0.333,0.481,0.521,0.236,0.791,0.187,0.819
snapshots:
enabled: true
record:
enabled: true
retain:
days: 14
mode: all
events:
retain:
default: 30
mode: motion
motion: {}
Front: # <--- name assigned
enabled: true
ffmpeg:
hwaccel_args: preset-vaapi # < --- leave this on as there will always be CPU/GPU use by frigate
inputs:
# high res stream
- path: rtsp://admin:[email protected]/Preview_01_main
roles:
- record
- audio # < --- this allows audio on exports/recordings
# low res stream
- path: rtsp://admin:[email protected]/Preview_01_sub
roles:
- detect
output_args:
record: preset-record-generic-audio-copy
detect:
enabled: false
width: 2560
height: 1920
fps: 5
objects:
track:
- person
filters:
person:
mask: 0.587,0.012,0.608,0.105,0.999,0.544,0.997,0.004
mask: 0.186,0.061,0.255,0.446,0.062,0.739,0.002,0.587,0.001,0.201
motion:
mask: 0.608,0.105,0.997,0.549,0.997,0.001,0.589,0.004
snapshots:
enabled: true
record:
enabled: true
retain:
days: 14
mode: all
events:
retain:
default: 30
mode: motion
version: 0.14Relevant log outputType
Timestamp
Tag
Message
info
2024-08-22 08:21:51
startup
Preparing Frigate...
info
2024-08-22 08:21:51
startup
Starting Frigate...
info
2024-08-22 08:21:53
frigate.app
Starting Frigate (0.14.0-da913d8)
info
2024-08-22 08:21:53
frigate.util.config
Checking if frigate config needs migration...
info
2024-08-22 08:21:53
frigate.util.config
frigate config does not need migration...
info
2024-08-22 08:21:53
frigate.util.services
Automatically detected vaapi hwaccel for video decoding
info
2024-08-22 08:21:53
peewee_migrate.logs
Starting migrations
info
2024-08-22 08:21:53
peewee_migrate.logs
There is nothing to migrate
info
2024-08-22 08:21:53
frigate.app
Recording process started: 406
info
2024-08-22 08:21:53
frigate.app
Recording process started: 407
info
2024-08-22 08:21:53
frigate.app
go2rtc process pid: 98
info
2024-08-22 08:21:53
detector.coral
Starting detection process: 434
info
2024-08-22 08:21:53
frigate.app
Output process started: 436
info
2024-08-22 08:21:53
frigate.detectors.plugins.edgetpu_tfl
Attempting to load TPU as pci
info
2024-08-22 08:21:53
frigate.detectors.plugins.edgetpu_tfl
TPU found
info
2024-08-22 08:21:53
frigate.app
Camera processor started for Drive: 463
info
2024-08-22 08:21:53
frigate.app
Camera processor started for Front: 464
info
2024-08-22 08:21:53
frigate.app
Capture process started for Drive: 465
info
2024-08-22 08:21:53
frigate.app
Capture process started for Front: 468Frigate statsNo response Operating systemOther Linux Install methodDocker Compose Object DetectorCoral Any other information that may be helpfulNo response |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 9 replies
-
|
Your |
Beta Was this translation helpful? Give feedback.
Your
detectresolution is very high. It does not need to match the resolution of your camera. If you lower it you should see a significant drop in CPU usage. Start by trying half of what you have now.