I am still getting parsing config errors after making the changes used online yml validators and come back correct format?
Error parsing config: 'NoneType' object has no attribute 'get'
* Starting nginx nginx
...done.
I have finally been able to get the container to load, the problems was the config.yml file which i complied again. The container loads the 4 camera images, however only 2 work correctly providing clips and events , the other 2 do not do either, considering the configs are the same for each camera.
The cameras are Reolink 410, the motion detected os working on the Reolink app.
Any idea what the problem can be ?
* Starting nginx nginx
...done.
frigate.app WARNING : Camera garage_driveway has rtmp enabled, but rtmp is not assigned to an input.
frigate.app WARNING : Camera front_driveway has rtmp enabled, but rtmp is not assigned to an input.
frigate.app WARNING : Camera back_garden has rtmp enabled, but rtmp is not assigned to an input.
frigate.app WARNING : Camera koi_pond has rtmp enabled, but rtmp is not assigned to an input.
Starting migrations
peewee_migrate INFO : Starting migrations
There is nothing to migrate
peewee_migrate INFO : There is nothing to migrate
detector.cpu1 INFO : Starting detection process: 34
detector.cpu2 INFO : Starting detection process: 35
frigate.app INFO : Camera processor started for garage_driveway: 39
frigate.app INFO : Camera processor started for front_driveway: 40
frigate.app INFO : Camera processor started for back_garden: 41
frigate.app INFO : Camera processor started for koi_pond: 42
frigate.app INFO : Capture process started for garage_driveway: 43
frigate.app INFO : Capture process started for front_driveway: 46
frigate.app INFO : Capture process started for back_garden: 50
detectors:
cpu1:
type: cpu
cpu2:
type: cpu
mqtt:
host: 192.168.0.74
user: mikey
password: mqtt6283
cameras:
# Reolink
garage_driveway:
ffmpeg:
inputs:
- path: rtsp://admin:[email protected]:554/h264Preview_01_sub
- roles:
- detect
- clips
# motion:
# mask:
width: 640
height: 480
fps: 5
objects:
track:
- person
snapshots:
enabled: true
timestamp: false
bounding_box: true
retain:
default: 1
clips:
enabled: true
retain:
default: 1
front_driveway:
ffmpeg:
inputs:
- path: rtsp://admin:[email protected]:554/h264Preview_01_sub
- roles:
- detect
- clips
# motion:
# mask:
width: 640
height: 480
fps: 5
objects:
track:
- person
snapshots:
enabled: true
timestamp: false
bounding_box: true
retain:
default: 1
clips:
enabled: true
retain:
default: 1
back_garden:
ffmpeg:
inputs:
- path: rtsp://admin:[email protected]:554/h264Preview_01_sub
roles:
- detect
- clips
width: 640
height: 480
fps: 5
objects:
track:
- person
snapshots:
enabled: true
timestamp: false
bounding_box: true
retain:
default: 1
clips:
enabled: true
retain:
default: 1
koi_pond:
ffmpeg:
inputs:
- path: rtsp://admin:[email protected]:554/h264Preview_01_sub
roles:
- detect
- clips
# motion:
# mask:
width: 640
height: 480
fps: 5
objects:
track:
- person
snapshots:
enabled: true
timestamp: false
bounding_box: true
retain:
default: 1
clips:
enabled: true
retain:
default: 1````Stupid question, but are the camera entities created in HA by the Frigate integration supposed to stream?
Itād be great to be able to utilise these directly without having to recreate them as generic cameras. I did try setting one camera with RTMP but could not figure out how to āseeā that stream. Three cameras are Hikvision and three are Dahua; all running the 3rd stream as 720P, h264 @ 6fps.
P.S Frigate is awesome, especially when you add and utilise the Frigate LL card. My wife is loving just being able to use the wall mounted tablet to have a quick look through all the 10s clips from the driveway and front door cameras (and the little image bubbles are genius!).
Yes, they are meant to stream. They use the RTMP feed that Frigate relays. Probably best to open an issue on Github for further support.
Thanks for the fast reply, much appreciated. And thatās great. I was hopeful they created a working camera as that would be bloody useful. Iāll pop something up in the morning as very late here in NZ.
I thought i had Frigate up and running but now i have this error and a lot of cpu useage ?
Any ideas ???
`
[h264 @ 0x55ac3e072e00] error while decoding MB 28 18, bytestream -11
[h264 @ 0x55ac3e072e00] left block unavailable for requested intra mode
[h264 @ 0x55ac3e072e00] error while decoding MB 0 5, bytestream 48387
[h264 @ 0x55b21cfe2e00] error while decoding MB 2 23, bytestream -55
[h264 @ 0x556109d3de00] error while decoding MB 33 3, bytestream -26
[h264 @ 0x556109d3de00] left block unavailable for requested intra4x4 mode -1
[h264 @ 0x556109d3de00] error while decoding MB 0 5, bytestream 45781
[h264 @ 0x563922e98e00] error while decoding MB 22 8, bytestream -16
[h264 @ 0x55d037e3ce00] error while decoding MB 22 2, bytestream -56
[h264 @ 0x55d037e3ce00] left block unavailable for requested intra mode
[h264 @ 0x55d037e3ce00] error while decoding MB 0 10, bytestream 65829
[h264 @ 0x55c8fb2d8e00] error while decoding MB 23 8, bytestream -20
[h264 @ 0x5615f966ce00] error while decoding MB 37 8, bytestream -37
[h264 @ 0x5615f966ce00] left block unavailable for requested intra mode
4 cameras, no TPU, no HW accel and only CPU. You canāt expect low overload and so frames are skipped
Thatās part of the documentation
It was working for a day and then all of a sudden it came up with these errors. I have ordered a Coral TPU but it will be a while yet. So i set this up in the meantime. I will try just 2 cameras to see if this solves the problem.
Same result after removing too cameras ??
So what is the point ?
If 1 cam is more than 50% CPU, 2 will be too muchā¦
No point, I am just trying to understand and learn what is going on. I will use HA Frigate addon as it works fine there with 4 cameras on a sub stream and NO TPU and I was using sub stream in the HA Container. I was trying to have the cameras separate from the rest of HA.
Thank anyway.
Ffmpeg is used to intially convert video to 300px by 300px. This will use a lot of cpu especially with no GPU
After 300x300, the TPU is used for detect object if you have itā¦if not also cpu used here
Thanks for reply, Is there a way that i can rectify this , as i donāt have a TPU as yet it was working fine for a few days then this . I followed the recommended configuration which includes the ffmpeg. if i were to remove it would this solve in the meantime? My understand now is that you have to use Coral for more than 1 camera ? If so I will have to wait until there are finally in stock ?
Hello! Someone connected a Sricam camera on frigate? I have the SP018 model. Iām never sure what path I should pass, how can I find this? Iām getting a green screen on frigate dashboard.
Turns out I did not have ffmpeg enabled in my config. I had it but Iād renamed the package file that contained that line while I had a camera offline. Put ffmpeg back into main config and streaming now. 
For anyone looking for PCIe to M.2 (e key)
oddly hard to find
wooohoo
Hi Everyone.
Iāve been running Frigate off a NUC that runs off an intel CPU.
It has been running stable for weeks with no issues, with cpu detection on 3 cameras.
I have also been using the object detection integration to trigger automations with no issue for a while.
Today, i decided to setup https, so i installed nginx for reverse proxy connections, and MariaDB (required by Nginx proxy manager). My reverse proxy is working fine, but weirdly, i noticed that my object detection was broken.
Iām not sure what broke it, when I disabled MariaDB and Nginx, as well as removed the configuration.yaml changes, the problem still persisted.
Iām getting these entries in my Frigate log.
* Starting nginx nginx
...done.
frigate.app INFO : Creating directory: /tmp/cache
Starting migrations
peewee_migrate INFO : Starting migrations
There is nothing to migrate
peewee_migrate INFO : There is nothing to migrate
frigate.mqtt INFO : MQTT connected
detector.cpu1 INFO : Starting detection process: 38
frigate.app INFO : Camera processor started for Dispensary_Front: 41
frigate.mqtt INFO : Turning off clips for Dispensary_Front via mqtt
frigate.mqtt INFO : Turning off snapshots for Dispensary_Front via mqtt
frigate.mqtt INFO : Turning off detection for Dispensary_Front via mqtt
frigate.mqtt INFO : Turning off clips for Waiting_Area via mqtt
frigate.mqtt INFO : Turning off snapshots for Waiting_Area via mqtt
frigate.mqtt INFO : Turning off detection for Waiting_Area via mqtt
frigate.mqtt INFO : Turning off clips for FiveFootWay via mqtt
frigate.mqtt INFO : Turning off snapshots for FiveFootWay via mqtt
frigate.mqtt INFO : Turning off detection for FiveFootWay via mqtt
frigate.app INFO : Camera processor started for Waiting_Area: 42
frigate.app INFO : Camera processor started for FiveFootWay: 43
frigate.app INFO : Capture process started for Dispensary_Front: 44
frigate.app INFO : Capture process started for Waiting_Area: 48
frigate.app INFO : Capture process started for FiveFootWay: 49
I canāt make sense of the mqtt info bit, but my presence detection isnāt working.
Any idea what I might have broke?
Thanks !
Omg new update birdseye omg waaaw
Iāve been using Tensorflow for a couple of years with GPU acceleration. Right now I have a GTX 1080 behind my deployment and it is scanning six camera feeds every 3-5 seconds and doing full-frame object detection.
Iām starting to get concerned over the lack of updates to tensorflow and the fact that stats are only showing eight active deployments (Iām one of them).
Does Frigate support GPU acceleration? I see that itās based on Tensorflow, so I assume the inherent Nvidia support is included. I have a Coral but my expectations far exceed its capabilities.
My requirements:
- Real-time multi-camera object detection (6 x 2560x1920 images)
- Full-frame (not zoned) detection
Iām currently considering upgrading my GTX 1080 to something faster. Iād really like to have updates in the 1-2 second range.