I would limit cpu and memory use for all your containers.
It rare but I’ve had some containers that get excessive and take all memory causing crash or system lag. To prevent this always best to limit use
EDIT
And 8GB is a lot so you need to look into that
Hi everyone,
I have setup my cameras with motion masks and zones, but I still get notifications triggered by events outside the zone even though I specifically specified it. Using the blueprint for notifications BTW.
But also I get lots of events triggered by shadows from trees, lights on the road…
Is there a way to overcome those false positive? Otherwise the notifications are pretty much useless
Regards,
Didi
are shadows,lights on the road identified as person?
alias: NOTIFY_FrigateEntryGateMotionInOut
description: ''
trigger:
- platform: mqtt
topic: frigate/events
condition:
- condition: or
conditions:
- condition: template
value_template: '{{ trigger.payload_json["after"]["current_zones"] == ["outside"] }}'
- condition: template
value_template: '{{ trigger.payload_json["after"]["entered_zones"] == ["outside"] }}'
- condition: template
value_template: '{{ trigger.payload_json["after"]["current_zones"] == ["inside"] }}'
- condition: template
value_template: '{{ trigger.payload_json["after"]["entered_zones"] == ["inside"] }}'
- condition: template
value_template: '{{ trigger.payload_json["after"]["camera"] == "Frigate_Gate02"}}'
action:
- service: notify.mobile_app_terrences_iphone
data:
message: >-
A {{trigger.payload_json["after"]["label"]}} is {{
trigger.payload_json["after"]["current_zones"] }} the
Gate.{{trigger.payload_json["after"]["top_score"]}}
data:
tag: NOTIFY_EntryGateMotionInOut
image: >-
https://haweb.com/api/frigate/notifications/{{
trigger.payload_json["after"]["id"] }}/snapshot.jpg
actions:
- action: OPEN_ENTRY_GATE
title: Open Gate
destructive: true
- action: URI
title: APP
uri: /lovelace/3
- delay:
hours: 0
minutes: 0
seconds: 40
milliseconds: 0
mode: single
@ukro no, they are on the car, so it detects it as the car but the car was standing / stationary. The shadows / reflected lights from the roard5by other cars are casted on the car’s side and then consider to be a movement, but it is a small portion of the whole vehicle. Is there a way to ignore those movements? The issue with the motion masks is that if I’ll try to block all of those shadows it will practically block the whole detection area, and also I can’t really predict exactly where the shadow or the reflecting lights will appear. My question is if there would be a motion but on the car, can it determine that the car was there already and this wasn’t an event for a “car”?
Or are there any other methods to troubleshoot this?
wuf that’s hard one. i have no idea, i think you need to show some screenshots and ask @blakeblackshear
@tmjpugh I’m trying to configure it right now, I noticed that you set up two zones and the automation is based on that. For my setup I don’t use zones for my frontdoor cam, should I divide it into two separate as well? (sorry if it sounds like a dumb questions, I don’t have much experience with Frigate)
Thanks!
Didi
For the past couple of days Frigate has decided to stop deleting the clips which don’t have events in them, my conifig is setup ‘retain_days=0’.
It’s been working as expected for weeks/months but just suddenly started doing it a couple of days ago. As far as I can remember I’ve made no changes to Frigate or HA. This is resulting in my disk becoming full and me needing to manually delete clips, you can see the spikes in the screenshot below:
I’m running:
Frigate: 2.3
Installed as an add-on to HA core-2022.2.0 on Home Assistant OS 7.2, I was running 2021.12.x just upgraded today to Feb release to see if it would fix it.
This has happened to me once in the past, however after a few days it went back to deleting the clips so I never got to the bottom off it. Nothing in the Frigate or HA logs to indicate an issue. Any suggestions?
Config below, 11 cameras so it’s quite noisy:
mqtt:
host: 192.168.1.166
port: 1883
user: mqttuser
password: mqttPassw0rd
topic_prefix: frigate
client_id: frigate
detectors:
coral:
type: edgetpu
device: usb
ffmpeg:
hwaccel_args:
- -hwaccel
- qsv
- -qsv_device
- /dev/dri/renderD128
birdseye:
enabled: False
width: 1280
height: 720
quality: 4
mode: objects
snapshots:
enabled: True
bounding_box: true
retain:
default: 5
record:
enabled: True
retain_days: 0
events:
retain:
default: 5
objects:
track:
- person
- cat
- mouse
- dog
- bird
cameras:
front_door:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/4E0jPd2vPAlxvTxd
roles:
- detect
- record
detect:
width: 1024
height: 576
fps: 5
motion:
mask:
- 1024,482,934,431,963,287,1024,238
- 423,0,429,153,237,214,87,206,0,150,0,0,174,0
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: true
crop: False
height: 500
deck:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/x84oJyVPCbJkM6K3
roles:
- detect
- record
detect:
width: 1280
height: 720
fps: 5
motion:
mask:
- 925,0,831,71,681,106,432,275,0,375,0,0
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
front_bins:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/1D7DHGoMfMtuiRPA
roles:
- detect
- record
detect:
width: 1280
height: 720
fps: 5
motion:
mask:
- 1280,385,387,172,0,138,0,0,1280,0
zones:
front_of_house:
coordinates: 1280,720,0,720,0,93,114,131,250,149,725,224,1280,353
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
pool2:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/H6yu5EC02nug8e8a
roles:
- detect
- record
detect:
width: 1280
height: 720
fps: 5
motion:
mask:
- 241,48,403,142,537,269,569,720,0,720,0,487,0,88
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
front_bbq:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/0vRR1mnUZGIYbmFc
roles:
- detect
- record
detect:
width: 1280
height: 720
fps: 5
motion:
mask:
- 1280,0,1280,180,1191,182,1182,238,701,308,0,487,0,0
zones:
front_of_house:
coordinates: 474,358,1171,222,1186,176,1280,180,1280,720,0,720,0,480
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
back_garden:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/Mc4bF1KgSdwbWRov
roles:
- detect
- record
detect:
width: 1280
height: 720
fps: 5
motion:
mask:
- 816,720,766,474,775,383,984,159,1120,138,1280,320,1280,720
- 0,0,1280,0,1079,259,731,259,602,252,244,359,186,468,0,588
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
garage:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/ywNuWiY0nfzantkr
roles:
- detect
- record
detect:
width: 1024
height: 576
fps: 5
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
door_bell:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/r1Fil4PuD5xYntVv
roles:
- detect
- record
detect:
width: 960
height: 720
fps: 5
motion:
mask:
- 0,60,960,59,960,361,849,380,854,408,553,435,347,446,120,459,118,431,0,421
zones:
front_of_house:
coordinates: 960,720,0,720,0,425,148,441,152,483,960,371
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
livingroom:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/Lncnic1vxoDvlvkw
roles:
detect:
width: 1024
height: 576
fps: 5
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
office:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/ihBjWrdkIC4LqEAg
roles:
detect:
width: 1024
height: 576
fps: 5
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
kitchen:
ffmpeg:
inputs:
- path: rtsp://192.168.1.1:7447/127xhqW0hRMAlRsG
roles:
detect:
width: 1024
height: 576
fps: 5
rtmp:
enabled: false
mqtt:
timestamp: False
bounding_box: False
crop: False
height: 500
Zones are not required
This camera points at busy highway. Zones detect my driveway (inside and outside). I use frigate as doorbell to detect waiting vehicle, delivery person and mail car.
This automation send image of event and include actionable notification button (app and open gate)
Open gate trigger separate automation that open gate. App opens app on phone so I may check live video.
Zone limit detection to prevent false alert in my case but you can remove if not needed.
Long press notification will show low res image zoomed larger. Single press notification opens app. There is delay at end to prevent new notification send for 40 seconds since the tag will allow new notification to overwrite old vs create new seperate notification
Can I use required_zones for a specific object? for example, I want the snapshots with required_zones but for one zone it would trigger an event for “person” and another required_zone would be for “car”?
I think that would solve my false positive, if I would divide my detection area into 3 zones, only when a car appears in all 3 I would consider it as a true event (so the shadows/reflections won’t appear in all 3 at the same time). and for a person, it could be in the whole area of the detection. maybe I’m not looking at this as it is supposed to behave, correct me if this is not possible, or there is a better approach.
Thanks!
I noticed a couple of other people reporting the same issue: [Support]: Retention not working - recording all the time instead · Issue #2723 · blakeblackshear/frigate (github.com)
No idea why but after several Frigate, HA and HAOS restarts clips started deleting again as expected.
Can I update to the beta 10.x from 0.9.4-26ae608 or will I hit a breaing change (again)
I don’t remember there being a breaking change but….I can’t remember😊
Hi,
Can anyone help me to understand how to move the cache on an external SSD? (Samsung_T5)
I moved the media (clips, record, db, etc) but I don’t understand how to do the same for the cache
volumes:
- /etc/localtime:/etc/localtime:ro
- ./config:/config
- /media/pi/Samsung_T5:/media
- type: tmpfs
# target: /tmp/cache
target: /media/cache
tmpfs:
size: 1000000000
regards
Does anyone know if there is a way to monitor the coral stats ?
utilization
tps
temp
etc
Got Frigate running as a container , awesome piece of software.
It’s doing person detection on my doorbell and pushing it to my HA.
Only there’s a 45 secs lag between person detect and my HA notifying me there’s a person in the yard.
believe this is thanks to my I5 3,2 ghz inference speed to high i guess…? around 60ms
got 2 cores set in config, would it make a difference if i add all 4 cores ?
and would a Google Coral , make the detect → notify more realtime ?
I was using command line sensor
I believe frigate added this in latest beta release
This does Not work for usb coral
Thanks
Do you know where to find it in the latest beta ? I’m running that
Does Frigate itself see the event quickly ?
Is it a notification to your phone ?
I have person detection’s very quick to my phone, even before switching to a coral, maybe 2 or 3 seconds