0.11.0 Beta 2

[blakeblackshear]

Major Changes for 0.11.0

Link to updated docs: https://deploy-preview-2829--frigate-docs.netlify.app/

BREAKING CHANGES

This version includes a significant rework of the docker build. This may have broken ffmpeg hardware acceleration for some hardware that was working previously. Please open an issue if this happens so we can try and track down what's missing.

Frigate+ Integration

With this release, users can easily send snapshots to Frigate+ for labeling directly from the Events view. Note that Frigate+ is still in development, and custom models are not yet available. There is no cost to create an account and submit labeled images. Making it simpler to submit images should accelerate the build out of a robust training set.

To enable the Frigate+ functionality, you need to obtain an API Key from your Frigate+ account on the settings page and set it via the PLUS_API_KEY environment variable. This has been added to the settings page for the Home Assistant Addon as well.

Here is the roadmap for Frigate+:

• Launch web version allowing manual upload of training examples and labeling
• Integrate submission in Frigate
• Launch Frigate optimized models
• Launch individually optimized models
• Integrate labeling and suggestions in Frigate

Updated Events View

• Speed. The events view has been completely revamped and should be much faster than it was before.
• Slightly larger thumbnails.
• You no longer have to refresh the page to see the latest events.
• You can "star" or "favorite" events to tell Frigate to keep associated recordings and the snapshot indefinitely.
• Sub labels have been added to allow setting arbitrary labels on events. For example, you can use deepstack to run facial recognition and update the event with the name of the person detected. (FEAT: Ability to set sub labels for specific events #2949)

Updated Recordings View

• Speed. The recordings view doesn't look different, but it has been completely overhauled. It should be much faster, and event data and thumbnails are lazy loaded to reduce the amount of data that is loaded when visiting the page.
• Smoother seeking and transitions. Re-renders of the player have been reduced and you should see smoother seek behavior when selecting events.
• 16x playback was added for browsers other than Firefox.

Updated Dockerbuild

• Frigate is now built on top of plain Debian containers and consolidated into a single multiarch image.

Experimental Timeline UI

Not recommended for use yet, but an alternative UI has been added. It has very limited browser/device support. #2830

Other changes

• MQTT sensors for any object type FEAT: MQTT support for all objects for each camera & zone #2908
• Update latest object endpoints to reference most recent event from database (you will no longer see black images for objects under the camera in live view) FEAT: Replace best jpg endpoint #2944
• Added a new ratio filter for object filters Add object filter ratio #2952
• Camera recordings always in sidebar Always show camera recordings in sidebar regardless of state #2846
• Enable and set mode per camera for birdseye Allow birdseye to be overridden at the camera level #3083
• Customize order and hide cameras from the UI FEAT: Ability to reorder & ability to hide Cameras in UI #2981
• Enable improve_contrast via MQTT Allow improve_contrast to be toggled via mqtt #3011
• Enable/disable motion detection independent of object detection via MQTT Ability to enable / disable motion detection via MQTT #3117
• MQTT topic for motion detection Send mqtt message when motion is detected #3152
• Batch deletes of recordings when more than 100,000 recordings need to be deleted
• Switch to jellyfin ffmpeg
• Improve cache handling for thumbnails and snapshots of completed events
• Fixes hwaccel issues for Raspberry Pi users
• Updates to debian 11 as a base image
• Consolidate Dockerfile to enable multiarch images
• Birdseye icno can now be customized FEAT: Ability to set custom birdseye icon and birdseye docs #2979
• Frigate now checks for a newer release on startup and exposes that in the stats endpoint Add latest frigate version to stats endpoints #3038
• Initial implementation of mypy and type checking
• Images from backend now have appropriate cache headers

Full Changelog: v0.10.0...v0.11.0-beta2

Image: blakeblackshear/frigate:0.11.0-beta2

---

This discussion was created from the release 0.11.0 Beta 2.

[ozett]

just to inform others. it took 5-6 Minutes for migration .
on an old intel-i5 /ubuntu / docker / pci-e coral

new events-ui looks great!
fast as hell now with the date-picker.

would love to see another icon than the star
for permanent storage=write protection=delete protection of events.
The star looks like some kind of Rating System...

first impression: GREAT!

[NickM-27]

In general the star is used as a favorite but I can see that too.

Delete protection for favorited events is a good idea

[NickM-27]

PR for protection of favorited events: #3225

[hacsjalano]

I tried annotating my first uploaded image, but I see there are a bunch of preset types right now. Are you willing to add any more? In particular, I'd like to add "trash truck" because I'd like to know when they come by the front of my house!

I should add, at least where I live, they have a very distinct look which should be possible to separate from other vehicles.

[hacsjalano]

Related to the above - since I uploaded the image to my account, but I couldn't find an appropriate label, is there an option to delete the image from my account since I don't want to tag it as any of the existing items?

[NickM-27]

I'd like to add "trash truck" because I'd like to know when they come by the front of my house!

This is where the individually optimized models come in, for objects that are distinct but not common for everyone (or look different). You can add examples of your own image.

I couldn't find an appropriate label, is there an option to delete the image from my account since I don't want to tag it as any of the existing items?

Confused why would you not want to tag any of the existing items? When uploading an image ALL labels that are there should be tagged so the model can be trained.

Right now car would be the appropriate label for that garbage truck.

[hacsjalano]

Won't labeling the trash truck as "car" mean that when I eventually train it with my own model for "trash truck" that it might be mistaken for "car", as I still plan on using the built-in "car" images?

[NickM-27]

That's a fair question.

I believe the individually optimized models will still start with training set from all contributions for the other objects and once you add this new object a specific model will be trained for your use case at which point it won't be confused by car, but I could certainly be wrong about that.

Also just in general, just like how currently Amazon, FedEx, UPS are for the logo and the entire vehicle is still marked as car, might be better to have the garbage truck be labeled as a car and then the specific part with the logo or other unique aspect be the part that marks it as a garbage_truck.

[hacsjalano]

Also, a related question about how Frigate Plus is going to work - are the model images in the pull-down (Amazon, Fedex, person) going to be available for non Frigate Plus subscriptions? Or will only the ones which are custom (my example, trash truck, specific people in my household) be considered "custom models"?

[NickM-27]

Not to speak for Blake, but per #1043 (comment) both the (Amazon, Fedex, ...) and the individually trained models will be a part of the paid service.

[blakeblackshear]

All models created and delivered through Frigate+ will require a subscription. This includes anything generally optimized for Frigate and user specific models. You will still be able to train your own models or use other tensorflow models available elsewhere.

[hacsjalano]

BTW - on the events view, red trash icon is really huge!

[esmoyer]

You beat me to it, literally just updated and it's back to normal. Thanks for the great work you guys!

[Daniel-dev22]

Has the addon update been pushed? Still not seeing it and I reloaded the supervisor and "checked for updates" as well.

[esmoyer]

@Danny2100 are you talking about the HA add on to view frigate within HA? There was no update for that, but it now works again with beta4 of Frigate 0.11.0

[NickM-27]

The ha addon for frigate itself needs to be updated, I don't think it's been yet https://github.com/blakeblackshear/frigate-hass-addons/commits/main

[Daniel-dev22]

Makes sense as to why I don't see it yet thank you.

[davidcoulson]

Is there an updated 'Frigate NVR Proxy' addon for HA? I was able to update my Frigate container to 0.11.0-beta2 and access it directly, but I just got a blank page when accessing it via Home Assistant side bar.

[blakeblackshear]

This will be fixed in the next beta.

[davidcoulson]

This will be fixed in the next beta.

I just updated to 0.11.0-beta2 with the 1.3 Proxy add-on and i still get a blank page. Do I need to update the proxy addon somehow?

[NickM-27]

@davidcoulson Please update to frigate beta04

[davidcoulson]

@davidcoulson Please update to frigate beta04

Perfect - Works great. Thanks!

[Minglarn]

Strange... All my cam streams is green... ffmpeg is crying out loud that conversion failed ect ect..
Works flawless in 0.10..

[NickM-27]

Is this the same as or different than #3223 ?

If different than please create an issue

[esmoyer]

Might be the same issue I have: #3223

Try what Nick recommended on the issue going into the container and seeing if it finds ffmpeg

I think they are working on it

[Minglarn]

Hum... FFMPEG is found and it's decoding as I can see them now in main view.
However.. the detect process is complaining for all my cams. One of them is green...
Have to investigate...

[Minglarn]

Find the cause...
Seems that my output_args is not working anymore...
Having this in output_args:
record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:a copy -c:v copy

When removing this line it works... But it it hammers my CPU with 15-20% more usage... And recordings is not working... Hum..

[shred86]

Just upgraded from stable and noticed my CPU usage is near 100%. I created an issue. This is on an Odyssey X86 Blue J4125 with Loryta (Dahua) IPC-T5442TM-AS cameras. It looks like the increased usage is from ffmpeg.

[spantobi]

I must be blind but where is the new experimental timeline UI? :)

[spantobi]

Ah! I see, thank you! Much appreciated :)

[spantobi]

Just wanted to tell you that the timeline works pretty well! The jumping back and forward is smooth. The play button doesn't work for me but it's a great start :)

[NickM-27]

Credit for the implementation of the timeline UI goes to @johnmarksilly 👍

[ozett]

ui:
  use_experimental: true

just an idea:
if the development for the new UI takes some betas,
would be lovely to switch within between new/classic within the web-UI itself.

without fiddling in manually in the config..

[esmoyer]

I'm really liking the new UI and function it brings. Aside from the play not working, it would be nice if the camera stream was a little bigger. There is a lot of wasted screen real estate. Otherwise, it's awesome being able to jump from event to event!

[DrSpaldo]

Is there a nvidia image for this one?

[NickM-27]

It's all in the same image, check out the docs in the 0.11 branch for how to get it working

[DrSpaldo]

It's all in the same image, check out the docs in the 0.11 branch for how to get it working

Nice, thanks, I missed that :) Now I just have to find the docs

[DrSpaldo]

I must be blind, can someone point me to the docs for the beta?

[blakeblackshear]

https://deploy-preview-2829--frigate-docs.netlify.app/

[DrSpaldo]

So, I am getting strange results. It appears that 4 (Amcrest & Annke - frontdoor, backdoor, garage, frontdriveway) of my cameras are working fine. These use RTSP.

Then 3 (Reolink brand - backyard, frontroad, sidegarage) are not. Originally these cameras used RTMP which worked as any other modes wouldn't work for me.

I have an nvidia card inside, nvidia-smi shows no ffmpeg processes within the docker console, however, on the main unraid console it shows 4 processes using ffmpeg within nvidia-smi.

config.txt

dockerlog.txt

I haven't been able to find the docs for this specific version, so apologies if it is the config and not the version/new container...

[ozett]

there must be something configured to get nvidia-smi inside/outside of container. maybe...
i found suggestion for this here NVIDIA/nvidia-docker#1460 (comment)

[Minglarn]

Is anyone using Reolink http with audio?
As soon as I set this:
record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:v copy -c:a aac

ffmpeg spits out errors...
This is working with 0.10.x ...

[NickM-27]

What errors are you seeing? That exact command is working for all my cameras on 0.11

[Minglarn]

This is the config:

ffmpeg:
  hwaccel_args:
    - -hwaccel
    - vaapi
    - -hwaccel_device
    - /dev/dri/renderD128
    - -hwaccel_output_format
    - yuv420p

  global_args:
    - -hide_banner
    - -loglevel
    - debug

  output_args:
    # Optional: output args for detect streams (default: shown below)
    detect: -f rawvideo -pix_fmt yuv420p -filter:v fps=fps=5

# MY Working record:
## v0.10.1
#    record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:a copy -c:v copy

## v011.0-beta2
    record: -f segment -segment_time 10 -segment_format mp4 -reset_timestamps 1 -strftime 1 -c:a copy -c:v copy

  input_args:
    - -avoid_negative_ts
    - make_zero
    - -fflags
    - nobuffer+genpts+discardcorrupt
    - -flags
    - low_delay
    - -strict
    - experimental
    - -analyzeduration
    - 1000M
    - -probesize
    - 1000M
    - -rw_timeout
    - "5000000"

This is from the logs....

frigate    | [2022-05-19 20:04:22] watchdog.frigate_parkering     ERROR   : Ffmpeg process crashed unexpectedly for frigate_parkering.
frigate    | [2022-05-19 20:04:22] watchdog.frigate_parkering     ERROR   : The following ffmpeg logs include the last 100 lines prior to exit.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Decode to surface 0x13.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 1
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Param buffer (type 0, 672 bytes) is 0.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Param buffer (type 1, 240 bytes) is 0x1.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Slice 0 param buffer (3128 bytes) is 0x2.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Slice 0 data buffer (23102 bytes) is 0x3.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Decode to surface 0x12.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] nal_unit_type: 5(IDR), nal_ref_idc: 3
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Param buffer (type 0, 672 bytes) is 0x3.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Param buffer (type 1, 240 bytes) is 0x2.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Slice 0 param buffer (3128 bytes) is 0x1.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Slice 0 data buffer (519357 bytes) is 0.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [h264 @ 0x55dc5d628b00] Decode to surface 0x13.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [AVHWFramesContext @ 0x55dc5d815fc0] Map surface 0x13.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [AVHWFramesContext @ 0x55dc5d815fc0] Unmap surface 0x13.
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [Parsed_fps_0 @ 0x55dc5e2d1d00] Setting 'fps' to value '5'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [Parsed_fps_0 @ 0x55dc5e2d1d00] fps=5/1
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] Setting 'video_size' to value '2560x1920'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] Setting 'pix_fmt' to value '0'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] Setting 'time_base' to value '1/1000'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] Setting 'pixel_aspect' to value '0/1'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] Setting 'frame_rate' to value '30/1'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [graph 0 input from stream 0:0 @ 0x55dc5e823b00] w:2560 h:1920 pixfmt:yuv420p tb:1/1000 fr:30/1 sar:0/1
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [scaler_out_1_0 @ 0x55dc5e0e34c0] Setting 'w' to value '2560'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [scaler_out_1_0 @ 0x55dc5e0e34c0] Setting 'h' to value '1920'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [scaler_out_1_0 @ 0x55dc5e0e34c0] Setting 'flags' to value 'bicubic'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [scaler_out_1_0 @ 0x55dc5e0e34c0] w:2560 h:1920 flags:'bicubic' interl:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [format @ 0x55dc5e2cf840] Setting 'pix_fmts' to value 'yuv420p'
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [AVFilterGraph @ 0x55dc5e01c7c0] query_formats: 5 queried, 4 merged, 0 already done, 0 delayed
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [scaler_out_1_0 @ 0x55dc5e0e34c0] w:2560 h:1920 fmt:yuv420p sar:0/1 -> w:2560 h:1920 fmt:yuv420p sar:0/1 flags:0x4
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [Parsed_fps_0 @ 0x55dc5e2d1d00] Read frame with in pts 1999, out pts 10
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:26272 pts_time:1.642 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:26272 pts_time:1.642 dts:26272 dts_time:1.642 -> pts:26272 pts_time:1.642 dts:26272 dts_time:1.642
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:27296 pts_time:1.706 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:27296 pts_time:1.706 dts:27296 dts_time:1.706 -> pts:27296 pts_time:1.706 dts:27296 dts_time:1.706
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:28320 pts_time:1.77 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:28320 pts_time:1.77 dts:28320 dts_time:1.77 -> pts:28320 pts_time:1.77 dts:28320 dts_time:1.77
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:29344 pts_time:1.834 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:29344 pts_time:1.834 dts:29344 dts_time:1.834 -> pts:29344 pts_time:1.834 dts:29344 dts_time:1.834
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:30368 pts_time:1.898 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:30368 pts_time:1.898 dts:30368 dts_time:1.898 -> pts:30368 pts_time:1.898 dts:30368 dts_time:1.898
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] segment:'/tmp/cache/frigate_parkering-20220519200405.mp4' starts with packet stream:1 pts:31392 pts_time:1.962 frame:0
frigate    | [2022-05-19 20:04:22] ffmpeg.frigate_parkering.detect ERROR   : [segment @ 0x55dc5d629040] stream:1 start_pts_time:0 pts:31392 pts_time:1.962 dts:31392 dts_time:1.962 -> pts:31392 pts_time:1.962 dts:31392 dts_time:1.962

[NickM-27]

Are you using the same stream for detect and record? Seems interesting that hanging the record output_args would affect the detect stream, but that could explain it

[Minglarn]

Yes, using the same stream for -detect and -record.
However, my other cams have the same error... And they are separate streams.

[NickM-27]

Looks like there is a similar issue, maybe add your own notes there: #3235

[DrSpaldo]

Anyone got the new beta working with nvidia & unraid?

[DrSpaldo]

Hasn't made a difference switching to the other version. The part I am not sure about though is what @blakeblackshear wrote here:

  deploy:    # <------------- Add this section
    resources:
      reservations:
        devices:
          - driver: nvidia
            count: 1
            capabilities: [gpu]

Not sure where I would put that in unraid

[NickM-27]

Hasn't made a difference switching to the other version. The part I am not sure about though is what @blakeblackshear wrote here:

  deploy:    # <------------- Add this section
    resources:
      reservations:
        devices:
          - driver: nvidia
            count: 1
            capabilities: [gpu]

Not sure where I would put that in unraid

This is for docker compose. Unraid uses docker command line so this doesn't apply. You just need to add the --gpus all to the docker command

[DrSpaldo]

ah ok, well it has me beat. I am using 4 ffmpeg processes in nvidia-smi for the main unraid console. But, none in the console of the frigate container. 4 cameras seem to be working, 3 aren't

[2022-05-21 22:33:42] frigate.video ERROR : frontroad: Unable to read frames from ffmpeg process.

[2022-05-21 22:33:42] frigate.video ERROR : frontroad: ffmpeg process is not running. exiting capture thread...

[2022-05-21 22:33:43] frigate.video ERROR : sidegarage: Unable to read frames from ffmpeg process.

[2022-05-21 22:33:43] frigate.video ERROR : sidegarage: ffmpeg process is not running. exiting capture thread...

[2022-05-21 22:33:47] watchdog.sidegarage ERROR : Ffmpeg process crashed unexpectedly for sidegarage.

[2022-05-21 22:33:47] watchdog.sidegarage ERROR : The following ffmpeg logs include the last 100 lines prior to exit.

[2022-05-21 22:33:47] ffmpeg.sidegarage.detect ERROR : [mp4 @ 0x5637c2eaa100] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63488 >= 640

[2022-05-21 22:33:47] ffmpeg.sidegarage.detect ERROR : av_interleaved_write_frame(): Invalid argument

[2022-05-21 22:33:47] watchdog.backyard ERROR : Ffmpeg process crashed unexpectedly for backyard.

[2022-05-21 22:33:47] watchdog.backyard ERROR : The following ffmpeg logs include the last 100 lines prior to exit.

[2022-05-21 22:33:47] ffmpeg.backyard.detect ERROR : [mp4 @ 0x5649d7fdd980] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63488 >= 640

[2022-05-21 22:33:47] ffmpeg.backyard.detect ERROR : av_interleaved_write_frame(): Invalid argument

[2022-05-21 22:33:47] ffmpeg.backyard.detect ERROR : [mp4 @ 0x5649d7fdd980] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63488 >= 1664

[2022-05-21 22:33:47] ffmpeg.backyard.detect ERROR : Error writing trailer of /tmp/cache/backyard-%Y%m%d%H%M%S.mp4: Invalid argument

[2022-05-21 22:33:47] watchdog.frontroad ERROR : Ffmpeg process crashed unexpectedly for frontroad.

[2022-05-21 22:33:47] watchdog.frontroad ERROR : The following ffmpeg logs include the last 100 lines prior to exit.

[2022-05-21 22:33:47] ffmpeg.frontroad.detect ERROR : [mp4 @ 0x561bccac3840] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63696 >= 848

[2022-05-21 22:33:47] ffmpeg.frontroad.detect ERROR : av_interleaved_write_frame(): Invalid argument

[2022-05-21 22:33:47] ffmpeg.frontroad.detect ERROR : [mp4 @ 0x561bccac3840] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63696 >= 1872

[2022-05-21 22:33:47] ffmpeg.frontroad.detect ERROR : Error writing trailer of /tmp/cache/frontroad-%Y%m%d%H%M%S.mp4: Invalid argument

[2022-05-21 22:33:52] frigate.video ERROR : backyard: Unable to read frames from ffmpeg process.

[2022-05-21 22:33:52] frigate.video ERROR : backyard: ffmpeg process is not running. exiting capture thread...

[2022-05-21 22:33:52] frigate.video ERROR : frontroad: Unable to read frames from ffmpeg process.

[2022-05-21 22:33:52] frigate.video ERROR : frontroad: ffmpeg process is not running. exiting capture thread...

[2022-05-21 22:33:53] frigate.video ERROR : sidegarage: Unable to read frames from ffmpeg process.

[2022-05-21 22:33:53] frigate.video ERROR : sidegarage: ffmpeg process is not running. exiting capture thread...

[2022-05-21 22:33:57] watchdog.sidegarage ERROR : Ffmpeg process crashed unexpectedly for sidegarage.

[2022-05-21 22:33:57] watchdog.sidegarage ERROR : The following ffmpeg logs include the last 100 lines prior to exit.

[2022-05-21 22:33:57] ffmpeg.sidegarage.detect ERROR : [mp4 @ 0x56009021c2c0] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 63488 >= 640

[2022-05-21 22:34:24] frigate.video ERROR : backyard: ffmpeg process is not running. exiting capture thread...

[NickM-27]

At this point I'd recommend creating an issue and filling out the template so we can see the config and everything

[yayitazale]

I'll try to test this to prepare the changes to the templates

[DrSpaldo]

Anyone know if it’s possible to switch back to the previous beta without causing any issues with the database?

[NickM-27]

You would need a backup of the db before doing the update. Reverting versions from 0.11 will not work with the new fields in the DB

[DrSpaldo]

So just start a new database then keep the old one for when I go back to this beta line?

[NickM-27]

Going from 0.10 there will be no issues. The only concern is if you didn't copy the db BEFORE updating to 0.11 then there will be issues going back.

If you're asking about saving the 0.11 db in case you want to come back later, I'm not sure that makes sense unless you are going to save all of your snapshots / recordings as well, just saving the db won't do anything since the linked snapshots and recordings will be deleted.

[markfrancisonly]

Thanks for making the best linux-based NVR solution available today 😊 Keep up the good work! Best wishes

[ajevremovic]

I noticed a strange bug with this version. When I try to access the recordings for some camera (e.g. http://server/recording/SomeCam1), and then I click on a specific hour (e.g. 18:00, http://server/recording/SomeCam1/2022-05-30/18) it will wait for some time loading, and then it will "fall back" to that day at 00:00. Or, in some cases, it will play ~10 seconds and then go back to 00:00.

It happens both with recordings made with the new and the old versions of Frigate.

[NickM-27]

Do you see any browsers errors if you click on a recording hour and then view logs in Inspect Element -> Console?

[ajevremovic]

Both in Google Chrome and Firefox I see this is the error: Uncaught (in promise) DOMException: The play() request was interrupted by a new load request. https://goo.gl/LdLk22

Before that (while it tries to load requested video) there are multiple errors like this:
GET http://server:5000/vod/2022-05/30/18/SomeCam1/seg-1-v1-a1.ts 503 (Service Temporarily Unavailable) VideoPlayer.c5945fb6.js:10 GET http://server:5000/vod/2022-05/30/18/SomeCam1/seg-1-v1-a1.ts 503 (Service Temporarily Unavailable)

There are no server side errors in the log, just managing/terminating socket.

Also, I see 503 (Service Temporarily Unavailable) very often, e.g. when trying to load some event.

[bm55]

The Events page no longer shows the thumbnail with bounding box, area information. Do I need to update my config to show the thumbnail with the bbox area info? OR should I request that feature -either thumbnail or the size info displayed on the page. This is very helpful to eliminate when false alerts are received with out having to go through the snapshot images stored in frigate/clips .

[NickM-27]

Currently this is the behavior, so would need a feature request issue for this option to be added.

[bm55]

#3281
Thanks

[Homas]

It's a very nice UX/UI upgrade. It is became very easy to delete events the only issue with that are possible misclicks. Are you considering to add a trash/recycle bin with a possibility to restore events or should I open a feature request?

[NickM-27]

In the next beta there will be delete protection for favorites events. If that's not suitable open a request issue

[Homas]

do u mean like in previous versions a popup window? If yes, I've already have other feature request to support mass delete. Is it in scope?

[NickM-27]

Right, it's a pop-up to confirm deletion if you try to delete a favorited event (star)

To me mass delete is less necessary with the one click delete.

Could still be done though, just throwing my 2ç

[Homas]

if the confirmation will be required fo the "stared" events when trash bin will be appreciated for the other events due to possible misclicks (I like it but concerned that my fat fingers delete something :).

[markfrancisonly]

been thinking about the favorite and delete buttons some more, and found the reason I fundamentally disagree with it ....

we are creating manual work for the end-user, why not automating tasks, instead of offering work? I can see how favoriting an event would be useful if you have a single camera, such as a doorbell, and some stranger arrives to your door and you want to save their face for future reference. favoriting seems pretty cool in this scenario. Also maybe your mistress came up to your door, and now you want to delete her face off of your server before your wife finds out. Great. these are active user management tasks, and require constant gardening to manage and clean up the favorited clips in your storage. again, in my not so humble opinion, as developers we should be pushing towards complete automation with machine learning.

now I've said all of this but what I express that I don't disagree with the option to preserving a recording or releasing unneeded storage space. I have had to save video recordings on more than one occasion manually and it is a pain: it requires that I go to the storage array, navigate the folder structure, copy the recorded files (10 seconds at a time), if more than an hour, go back into the folder structure and deal with duplicate file names, then manually run a ffmpeg script to combine the files back into a single video file.

Fine, favoriting could help, but the object event timeline has never been complete enough for my recording preservation needs, and so far, being able to favorite an event wouldn't have helped me. I need to favorite an entire video recording time range, and eventually download that favorited time range out to save in archive class storage or delete it

Still can you have someone do this for me?

[NickM-27]

I don't really understand the initial point of view. Seems like you say you fundamentally disagree with the feature, explain scenarios where it would be useful, and then express desire for a similar feature in recordings.

To me "fundamentally disagree with" implies you think it shouldn't exist at least in its current implementation, but the rest of what you said doesn't align with that so perhaps I misunderstood.

I'm sure you know everyone uses frigate differently so naturally it is very useful for others. Favoriting / retaining events was a very highly requested feature and I've personally used it a decent amount saving clips of my dog doing funny stuff in the backyard and also someone we didn't expect walking around the front yard.

Also as far as automating, that's already been discussed and is coming, but gotta start somewhere and I see no reason for manual favoriting to not exist.

As far as recordings go, there's been a number of requests for improved management and information with recordings and it is definitely on the roadmap.

[Homas]

IMO automation is the next step after the manual management. E.g. I've 16 cameras and can not keep the whole recordings more for than 3 days but I do want to keep events for 90 days (e.g. to help police with crime investigations, they have a local program for that). I'm checking the events once in a while (because I send all of them and important ones to telegram channels) and when I do so, I want to cleanup the events obviously I don't need (e.g. when I mow my lawn).
I like the favourite function and just want to make "mass delete" a bit easy with a protection against "fat fingers".

[markfrancisonly]

this morning the baby was asleep so I had too much time, but I will elaborate with two very different usage scenarios (which we both know overlap, sarcasm intended):

just having fun

• machine learning, cool, I can save a clip of my neighbors cat!
• look mom, I can save and delete videos
• device crashes and burns, there's a lot to learn with this new technology, better than paying the man
• trigger fart sound when downstream process recognizes son's face
• favorite event of dog doing a trick on camera to show my bae (must delete later or my surveillance drive will fill up, no big deal)

use for utility

• machine learning used to recognize cars, estimate speed and record license plates
• create reliable record of persons coming and going, delete for any other purpose other than disk overload undesirable
• runs without crashing, storage doesn't overfill and corrupt system because otherwise solution doesn't have utility
• my god, I've spent too much time setting this up. glad it's up and running and I can do something else
• download and share event of person's face who broke into my house and suspected in a home invasion down the street

[markfrancisonly]

[ozett]

this statistic-table is calling for a FR on the debug page... 👀

[markfrancisonly]

source -

select camera, label, round(avg(count),0) as average_per_day from 
(
    select camera, label, round(start_time/86400,0) as day, count(round(start_time/86400,0)) as count from event
    group by camera, label, round(start_time/86400,0)
) a 
group by camera, label
order by average_per_day desc;

[LaurenceGough]

It was a real shame this broke the HA Addon, a bit of a nasty surprise when I woke up, and reverting back corrupted the Frigate DB so I lost all events and detection was broken :(. However, after spending many hours converting my whole HA setup to standalone containers to allow for this new version to run in its own container I must say the Frigate 11 Beta 2 build is incredibly fast and responsive when viewing events and seeking through them, super nice work there! This was an area that was a struggle before so it's super sweet to see this completely resolved, thank you!

Unfortunately this then prevented easy external access to the Frigate UI via Home Assistant so I'm trying to deal with that now, the HA camera streams via the Frigate card are just no comparison to the Frigate UI in reliability, ease of use (zooming etc, etc) and performance no matter what stream option is used.

Thank you for your hard work on this.

[messnerdev]

Excited about the Events UI revamp. I'm still on 0.10 and my busier camera won't even load the events page anymore. Just spins for hours.

[BeardedTek]

@blakeblackshear I haven't dove into anything quite yet along this line, but is there an API to send snapshots to frigate+? I'd like to integrate this into fEVR as well. The more sources providing input the better, right?
Love the improvements you're making!

[NickM-27]

It's here https://github.com/blakeblackshear/frigate/blob/release-0.11.0/frigate/http.py#L137

[BeardedTek]

Thank you @NickM-27!

[markfrancisonly]

amazing impetus and wise use of energy. the performance gains brought to v11 are a major improvement!

please consider moving the favorite button off the event thumbnail

[Homas]

Not sure if it is an old bug or a new one. In the recordings, when you click to watch a recording for a specific hour and it start playing the following hour, when you click the previous hour (you started watching) - nothing happens, the player doesn't start playing it again.
E.g. if you click on 9am, the player starts with 9am recordings and after that automatically continues showing 10am recordings than if you want to watch 9am recordings and click on 9am - nothing happens. So to navigate back you will need to click on any other hour including 10am and navigate back.

[NickM-27]

If you inspect element and check the console do you see any browser errors?

[Homas]

No errors in console. The element looks ok. May be it is in issue, because browsers thinks that it is the same page?
For debug I used Chrome, but usually I'm using Safari due to h.265

[blakeblackshear]

I can reproduce this and I know why it happens. It is due to the url not changing. I need to think about the best way to address it.

[markfrancisonly]

serious question, where to begin deleting events? want to keep the max full-day recordings I can afford, but at least 30 days of events

record:
  enabled: True
  retain:
    days: <MAX # OF DAYS I CAN AFFORD>
    mode: all

  events:
    retain:
      # retention days
      default: 30
      mode: active_objects

right now event count > 100,000

select count(*) from event

and this is happening

[2022-06-12 16:13:33] frigate.record                 ERROR   : [Errno 28] No space left on device: '/media/frigate/recordings/2022-06/12/16/kitchen'
[2022-06-12 16:13:37] frigate.record                 ERROR   : Error occurred when attempting to maintain recording cache
[2022-06-12 16:13:37] frigate.record                 ERROR   : [Errno 28] No space left on device: '/media/frigate/recordings/2022-06/12/16/kitchen'
[2022-06-12 16:13:42] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/baby-20220612161327.mp4
[2022-06-12 16:13:42] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/baby-20220612161327.mp4' -> '/media/frigate/recordings/2022-06/12/16/baby/13.27.mp4'
[2022-06-12 16:13:42] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/back_patio-20220612161328.mp4
[2022-06-12 16:13:42] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/back_patio-20220612161328.mp4' -> '/media/frigate/recordings/2022-06/12/16/back_patio/13.28.mp4'
[2022-06-12 16:13:42] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/back_windows-20220612161328.mp4
[2022-06-12 16:13:42] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/back_windows-20220612161328.mp4' -> '/media/frigate/recordings/2022-06/12/16/back_windows/13.28.mp4'
[2022-06-12 16:13:43] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/back_yard-20220612161329.mp4
[2022-06-12 16:13:43] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/back_yard-20220612161329.mp4' -> '/media/frigate/recordings/2022-06/12/16/back_yard/13.29.mp4'
[2022-06-12 16:13:43] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/doorbell-20220612161328.mp4
[2022-06-12 16:13:43] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/doorbell-20220612161328.mp4' -> '/media/frigate/recordings/2022-06/12/16/doorbell/13.28.mp4'
[2022-06-12 16:13:43] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/family-20220612161328.mp4
[2022-06-12 16:13:43] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/family-20220612161328.mp4' -> '/media/frigate/recordings/2022-06/12/16/family/13.28.mp4'
[2022-06-12 16:13:43] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/front_door-20220612161328.mp4
[2022-06-12 16:13:43] frigate.record                 ERROR   : [Errno 28] No space left on device: '/tmp/cache/front_door-20220612161328.mp4' -> '/media/frigate/recordings/2022-06/12/16/front_door/13.28.mp4'
[2022-06-12 16:13:43] frigate.record                 ERROR   : Unable to store recording segment /tmp/cache/front_street-20220612161328.mp4

queried the event db table for answers:

select  id, label, camera, DATETIME(start_time, 'unixepoch') as start_datetime,  DATETIME(end_time, 'unixepoch')  as end_datetime,(round(end_time / 3600,0) - round(start_time / 3600,0)) as duration from event where (round(end_time / 3600,0) - round(start_time / 3600,0)) > 2 order by (round(end_time / 3600,0) - round(start_time / 3600,0)) desc;

id	label	camera	start_datetime	end_datetime	duration in hrs
1652564390.582846-1pa068	car	front_street	5/14/2022 21:39	5/17/2022 14:36	65
1653418857.335469-91nau9	car	front_street	5/24/2022 19:00	5/26/2022 11:58	41
1650451958.629805-x4z045	person	baby	4/20/2022 10:52	4/22/2022 1:47	39
1650507498.359847-5hlhsf	person	family	4/21/2022 2:18	4/22/2022 13:20	35
1654643380.08123-8owipg	car	front_street	6/7/2022 23:09	6/9/2022 0:20	25
1653266023.512303-vnl6lg	car	front_street	5/23/2022 0:33	5/23/2022 21:37	21
1653349527.899697-cxt2qf	car	front_street	5/23/2022 23:45	5/24/2022 18:51	19
1654734681.876367-l0adlz	car	front_street	6/9/2022 0:31	6/9/2022 18:50	18
1652692654.385533-uejf8w	car	front_yard	5/16/2022 9:17	5/17/2022 1:37	17
1652910824.185046-tr21ht	car	front_yard	5/18/2022 21:53	5/19/2022 14:34	17

... found 400 events with a duration of more than 2 hours.

I can calculate a single full-day recording storage requirement by retrieving the folder size for a complete day

maximum # retained recorded days= drive size / full-day folder size

but events don't fit the formula. events are favorited or not yet expired, how do I clean up events from a 100,000 item list? and how do I prevent this from reoccurring?

[NickM-27]

Events use the same storage that the full recording segments use, so there's no overlap there for the days that are fully recorded.

Typically the segment sizes are similar so can be calculated by that. See #2790 (comment) for an example on how to calculate this.

[markfrancisonly]

Events use the same storage that the full recording segments use, so there's no overlap there.

using the following settings, are 30 days of events clips retained or only 3?

record:
  enabled: True
  retain:
    days:  3
    mode: all

  events:
    retain:
      # retention days
      default: 30
      mode: active_objects

say I can afford storage for 3 full days and have 20% storage free for events clips. want to retain 30 days of event clips, and okay delete clips using first-in-first-out overwriting queue. clean up can be performed automatically, or manually, but don't want the nvr to stop recording due to retained events

[NickM-27]

3 days of full recording segments would be kept which would be the same segments used for clips. Then 27 days of event clips would be kept.

say I can afford storage for 3 full days and have 20% storage free for events clips. want to retain 30 days of event clips, and okay delete clips using first-in-first-out overwriting queue.

Right, that makes sense.

clean up can be performed automatically, or manually, but don't want the nvr to stop recording due to retained events

As of now it will stop if it runs out of storage, like has been mentioned there is already the FR to have an option for retention to be based on storage

[markfrancisonly]

simply adding up the event clip durations in my database - edited: filtered by has_clip

select  round(sum((round(end_time / 60,0) - round(start_time / 60,0))) / 60 / 24,0) as 'full-days of event clips' from event where has_clip = 1 

yields 321 120 days, which is many times greater my maximum full-days drive capacity for 16 cameras. this is an overstatement of the problem because events overlap? nonetheless my frigate instance is configured to crash in the future due to event retention disk overload

It seems that the only reliable way to prevent these settings from crashing the system is to set event retention days equal or less than max full-day recording capacity. if using favorites to permanently retain events, event retained days must be lower than full-day recording capacity in v.11

[Daniel-dev22]

I set the frigate plus environmental variable and I keep getting this error

[2022-06-14 14:05:27] frigate.http                   ERROR   : PLUS_API_KEY environment variable is not set

I have this in my configuration.

environment_vars:
 PLUS_API_KEY: apikey

[NickM-27]

Thanks, I was able to reproduce this and have a fix in #3344

[Daniel-dev22]

Just making sure it wasn't me lol. I followed the steps correctly?

[NickM-27]

Yes it is a bug

[blakeblackshear]

Setting the env variable in the config is intentionally unsupported. I don't want a bunch of API keys in support requests.

[esmoyer]

Has anyone else experienced recordings jumping back to "midnight" or the beginning of the day when either scrubbing (jumping to another place on the playback progress bar) or just randomly when it's showing the circle to load the next 10 second clip? I was searching through a recording of the 5am hour and jumping to different spots when after a couple clicks of it jumping 5:10am, 5:14am, it suddenly just went back to 12:00am. I was then just trying to watch the recording and it would play fine (get the loading circle every 10 seconds or so) but after a bit, instead of loading the next 10 second segment, it jumps back to midnight.

I can grab logs to open an issue, just not sure what would be most useful in this case since it's playback.

[NickM-27]

So, just to confirm, is it only when frigate loses focus or does this occur when seeking and frigate has focus the entire time?

[esmoyer]

Correct. I just checked it again and click to jump around the 5 o'clock hour over a dozen times. I think because I get the spinning circle that sometimes takes a few seconds I would go to my other screen to do something while it loaded, not even considering taking focus away caused it. Then it would load and start playing but when I would click back to the frigate window, it would reset to midnight.

I can grab logs, just let me know what would be most helpful in this instance.

[NickM-27]

I am able to reproduce it on my test setup. If I am watching recordings from a day that only has one hour it does not happen, only for days that have multiple hours.

Seems that it is resetting to the first item in the playlist only when there are multiple items in that playlist.

[NickM-27]

Also I don't think there will be any logs for this, I see no errors or anything.

[blakeblackshear]

This happens because SWC fetches new data on focus. When that happens the player rerender and starts at the time specified in the url. Same as the other issue.

[markfrancisonly]

noticed that segment_time other than 10-seconds doesn't seem to be supported in beta 4, discussing here #3406