0.15.0 Beta 1 #14653
Replies: 106 comments 397 replies
-
|
impressive pace over the last months! Frigate is now the best NVR tool |
Beta Was this translation helpful? Give feedback.
-
|
Thank you so much for the hard work the Frigate team! I just upgraded, yes I've been stalking this upgrade for a while now and see a few issues. Seeing the following errors occasionally in the logs and didn't see anything like this in 0.14.1: When I look at the cameras they are fine. The cameras reporting this are all Amcrests and I run all my cameras through go2rtc and Frigate consumes those of course via go2rtc. Going to the review section when I hover over some of the events, seems to be the most recent ones, I see a blank square but others show the mini preview image. 
Looking forward to helping out and making this best possible. Let me know if you need anything else to help troubleshoot. |
Beta Was this translation helpful? Give feedback.
-
13th Gen Intel(R) Core(TM) i7-13700 |
Beta Was this translation helpful? Give feedback.
-
|
You need to use qsv |
Beta Was this translation helpful? Give feedback.
-
So change: to ? |
Beta Was this translation helpful? Give feedback.
-
|
Depends what codec your detect streams are, but in general yes |
Beta Was this translation helpful? Give feedback.
-
|
Oh see I thought those settings were always related to the record stream not detect. My cameras are configured with main stream as 265 and sub-stream (detect) for 264. I made the qsv 264 tweak and will watch now for issues. |
Beta Was this translation helpful? Give feedback.
-
|
Fantastic progress as ever! One point on Semantic Search - interested as to where the 16GB RAM requirement comes from for the CLIP model. I've been running 0.15 dev builds for a few weeks and my Frigate container (which is set to 16GB max RAM on a 32GB system) has never exceeded 4.5GB memory use. I also only have an Intel 10700 iGPU and don't see any real performance problems (I think it took about 20 minutes to re-encode 7,000 snapshots). Appreciate you might not want to lower the requirement, but just sharing my experience to avoid putting anyone else off... Of course, 0.16 sounds like a GPU could become more mandatory for face recognition, so whether the long term view is that it's necessary would be a fair one. |
Beta Was this translation helpful? Give feedback.
-
This is part of why I see merit in a roadmap of planned features (not a timeline, but a feature list) - so users can make the right hardware choices. I've been pondering this a bit so will make a separate post on the subject. |
Beta Was this translation helpful? Give feedback.
-
|
The thing is it is not really planned, and the supported hardware is also not known. The face recognition / alpr was not planned for 0.16 until we started talking about it literally a week ago. And it was not clear what hardware would be supported until we actually tried it out, found a model, etc. Things could certainly still change with it too. |
Beta Was this translation helpful? Give feedback.
-
|
As we've indicated elsewhere, there is no roadmap, no official development team, and no funding for Frigate beyond what users give via Github Sponsors. Frigate+ is exclusively Blake's project, Nick and I don't contribute or receive benefit from it. Nick and I are just volunteer contributors, and like he said, it was just last week that both of us worked on separate projects (face detection and license plate recognition) and decided to contribute them to the Frigate codebase. For a list of planned features, you can always take a look at the pinned issues. These are things users have requested that we feel are worthwhile to eventually have within Frigate. |
Beta Was this translation helpful? Give feedback.
-
|
I guess I'm as much asking from the perspective that you guys are probably as well-informed as anybody on the deployment of AI models relevant to the security space! Also appreciate that you're not sat there with the equipment or time to test across loads of different hardware configurations. Perhaps I could create a simple "benchmark.py" script that can be run as a non-default docker image command by Frigate, or a UI button, that executes a set of AI activities against some pre-supplied images and outputs the time taken |
Beta Was this translation helpful? Give feedback.
-
|
it's possible things like semantic search can have inference statistics just like the object detectors as a way to gauge how things are running. |
Beta Was this translation helpful? Give feedback.
-
|
Is there an update to https://github.com/blakeblackshear/frigate-hass-integration or https://github.com/dermotduffy/frigate-hass-card required? I can see the clips in the card, but when I click one to play it just shows a loading spinner and doesn't play any clip. |
Beta Was this translation helpful? Give feedback.
-
|
No, nothing has changed in these regards |
Beta Was this translation helpful? Give feedback.
-
|
Do you have plans to accept multiple languages? |
Beta Was this translation helpful? Give feedback.
-
|
It would likely make sense at some point, that would be a very large effort though to be clear |
Beta Was this translation helpful? Give feedback.
-
Thank you to the team's hard work and dedication. |
Beta Was this translation helpful? Give feedback.
-
|
First of all thank you to the team (and contributors) for all the time spent making Frigate so incredible! I was having some issues after upgrading, related to having birdseye enabled with restream. Looks like as I don't have an ffmpeg>hwaccel_args config key, Seems like amending this line to the below (adding
|
Beta Was this translation helpful? Give feedback.
-
|
auto as a default doesn't really make sense because the go2rtc config is created before frigate starts. Realistically you will need to set the hwaccel_args globally to ensure birdseye is restreamed efficiently. But yes setting the default as an empty string there will solve the issue so feel free to put up a PR |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the quick response, raised under #14658 |
Beta Was this translation helpful? Give feedback.
-
|
Found another issue related to Tracked Object Details interface, the left and right arrows at the bottom don't do anything so having to click on the icons in the middle section works but not the left or right:
|
Beta Was this translation helpful? Give feedback.
-
To be honest that is kind of confusing, should they be hidden if large enough or just active them for a more linear progression? |
Beta Was this translation helpful? Give feedback.
-
|
I'm only explaining that the behavior appears to be expected and not a bug in the buttons. Hiding them might be preferable |
Beta Was this translation helpful? Give feedback.
-
|
This is using two carousel UI elements - the first is the one your camera frame is displayed in, which can be dragged on desktop and swiped on mobile. The second carousel is the one with multiple squares/icons. The arrows correspond to that carousel and will be active/not grayed out when there are many steps in the object's life cycle. I can see how this may be confusing. We'll improve it for the next beta. |
Beta Was this translation helpful? Give feedback.
-
|
Just put up a PR to cause the buttons to move through the object lifecycle elements #14662 |
Beta Was this translation helpful? Give feedback.
-
|
I think it might be less confusing if the arrows are right next to the icons, rather than being spread wide. |
Beta Was this translation helpful? Give feedback.
-
|
So upgrading from 0.14.1 to this, for the first 10 minutes or so, the cameras seemed to load and then keep crashing over and over again. It seems to have stopped doing it for the next 10 minutes, so not sure yet if its just a temporary problem on startup/first upgrade, I can keep you updated... here's logs during the issue (we got the old "Unable to keep up with recording segments in cache for . Keeping the 6 most recent segments out of 8 and discarding the rest...": I haven't yet tried the new features, this was just a simple upgrade |
Beta Was this translation helpful? Give feedback.
-
|
I also noticed this with a few cameras. I restarted Frigate after about 10 minutes and it hasn't come back. I've also restarted it a few more times since then and it hasn't happened again. |
Beta Was this translation helpful? Give feedback.
-
|
Spun up my container using the beta and now i don't have the icon for frigate+ any longer on the left side of my screen. |
Beta Was this translation helpful? Give feedback.
-
Would it be possible to have an icon or any other visual effect on the explore view to see if an item was already sent to frigate+? We have to open a review item and go to the snapshot to see if it was already sent or if we can send it, could be good to know it without having to open it. |
Beta Was this translation helpful? Give feedback.
-
|
The send to plus button shows on the card if the object has not been sent and has a snapshot. You can also use the 3 dots |
Beta Was this translation helpful? Give feedback.
-
|
I do appreciate the new explore view a lot with all the features packed into it. As for Frigate+, let me know if I'm missing something (just installed 0.15 beta 1 today) but I do feel like it was much easier to skim through the snapshots before in the dedicated view, where the color coded bounding boxes would immediately tell me where to look, what was detected and the score. It was quick to spot a false positive, or badly placed bounding box, for example. In the new explore view I get something close to the old view when in Unfiltered Grid, but I need to look at the small icon to know what was detected and the score, and the thumbnail doesn't show the bounding box. It can happen that, for example, the icon says " 85%", there's a dog in the thumbnail, but it was actually a false positive in another part of the image. I would not suspect this false positive and would not open it to verify. I do miss some color coding, bigger confidence score/percentage and immediately viewable bounding boxes. The missing button as an indication that an imagine has been sent is also hard to quickly spot. I feel like this UI serves its main purpose, but not the Frigate+ purpose, as it's a lot harder to mentally filter through the thumbnails. Perhaps a dedicated Frigate+ view is still valid. Thanks a lot! Happy to continue testing the 0.15 betas. |
Beta Was this translation helpful? Give feedback.
-
|
In addition to the score filter, there have been some small changes to recent dev versions that will improve the Frigate+ workflow, including a "Submitted to Frigate+" filter for snapshots. After filtering, open up an item in Explore's grid view and navigate to the Snapshot tab where you'll see the familiar snapshot with the bounding box. The left/right arrow keys on your keyboard will move between each tracked object's snapshot for quick review and submission to Frigate+. This filter is available in the latest dev build and will also be in the next beta. |
Beta Was this translation helpful? Give feedback.
-
|
I did test the left/right arrow keys to move between snapshots but thought it was slow, as before I would just visually skim through many of those while scrolling. The new filter will help, and using other filters more aggressively too. I guess it takes some adjusting to the new workflow. Will keep an eye out for the next beta. Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
great to see new release!!! is it in my main config under each camera stream I need to add? |
Beta Was this translation helpful? Give feedback.
-
|
ok now I get the following error for camera "durys" (reolink doorbell) I double checked streaming from doorbell is using h264 |
Beta Was this translation helpful? Give feedback.
-
|
try using preset-vaapi for that camera |
Beta Was this translation helpful? Give feedback.
-
|
I changed global ffmpeg setting and error is gone. |
Beta Was this translation helpful? Give feedback.
-
|
if there are no errors then there should be no issue |
Beta Was this translation helpful? Give feedback.
-
|
anyway big thanks Nick. You are the best!!! |
Beta Was this translation helpful? Give feedback.
-
|
Hmm, after fixing the birdseye restream ffmpeg config I now hit this: |
Beta Was this translation helpful? Give feedback.
-
|
that would seem to indicate yaml is unable to load your config for some reason, perhaps it is invalid yaml |
Beta Was this translation helpful? Give feedback.
-
|
Fixed! My file was called now a new issue, all streams show perfectly in go2rtc, but not in frigate... |
Beta Was this translation helpful? Give feedback.
-
|
you should create a support discussion, but none of this seems related to the beta |
Beta Was this translation helpful? Give feedback.
-
|
That's weird because everything worked perfectly on 0.14.1 go2rtc:
streams:
tablet1: rtsp://10.9.8.137:8080/h264_pcm.sdp
tablet2: rtsp://10.9.8.109:8080/h264_pcm.sdp
tablet3: rtsp://10.9.8.64:8080/h264_pcm.sdp
tablet4: rtsp://10.9.8.163:8080/h264_pcm.sdp
audio: # <- enable audio events for all camera
enabled: true
snapshots:
enabled: true
cameras:
tablet1:
mqtt:
timestamp: false
bounding_box: false
crop: true
quality: 100
height: 500
ffmpeg:
inputs:
- path: "rtsp://127.0.0.1:8554/tablet1?video&audio"
input_args: preset-rtsp-restream
roles:
- audio
- detect
- record
tablet2:
mqtt:
timestamp: false
bounding_box: false
crop: true
quality: 100
height: 500
ffmpeg:
inputs:
- path: "rtsp://127.0.0.1:8554/tablet2?video&audio"
input_args: preset-rtsp-restream
roles:
- record
- detect
- audio
tablet3:
mqtt:
timestamp: false
bounding_box: false
crop: true
quality: 100
height: 500
ffmpeg:
inputs:
- path: "rtsp://127.0.0.1:8554/tablet3?video&audio"
input_args: preset-rtsp-restream
roles:
- record
- detect
- audio
tablet4:
mqtt:
timestamp: false
bounding_box: false
crop: true
quality: 100
height: 500
ffmpeg:
inputs:
- path: "rtsp://127.0.0.1:8554/tablet4?video&audio"
input_args: preset-rtsp-restream
roles:
- record
- detect
- audio
detectors:
coral:
type: edgetpu
device: usb
database:
path: /media/frigate/frigate.db
birdseye:
enabled: true
restream: true
mode: motion
motion:
threshold: 100
contour_area: 30
objects:
track:
- person
- face
- cat
ffmpeg:
hwaccel_args: preset-intel-qsv-h264
record:
enabled: true
expire_interval: 60
retain:
days: 1
mode: motion
alerts:
retain:
days: 7
detections:
retain:
days: 7
version: 0.15-0 |
Beta Was this translation helpful? Give feedback.
-
|
Fixed by setting ffmpeg section to: ffmpeg:
hwaccel_args: '""' |
Beta Was this translation helpful? Give feedback.
-
|
Impressive update, congratulations! I think I found a bug for the annotation offset. Otherwise, two suggestions:
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the report
|
Beta Was this translation helpful? Give feedback.
-
Thanks, fixed in #14667 |
Beta Was this translation helpful? Give feedback.
-
|
Great, thank you for the heads up and the fix! |
Beta Was this translation helpful? Give feedback.
-
|
Was there a breaking change? I did the update and tried to start Frigate and now I'm getting the following within the logs. I update almost every time there is one so I had whatever was the most recent before this morning. |
Beta Was this translation helpful? Give feedback.
-
|
This is a known issue that will be fixed for the next beta. |
Beta Was this translation helpful? Give feedback.
-
|
Ok, I will restore and ignore for now. :) |
Beta Was this translation helpful? Give feedback.
-
|
I had this issue and that was because I had Frigate+ API key configured in both config file and as environment variable. |
Beta Was this translation helpful? Give feedback.
-
|
Ok so semantic search is really cool. |
Beta Was this translation helpful? Give feedback.
-
|
The Home/End keys are bound to the suggestions list that appears below when entering filters through the keyboard. It does seem reasonable to have the keys work in the input when you're typing a normal search term without any filters in the input, though. I'll put up a PR to add this to the next beta. |
Beta Was this translation helpful? Give feedback.
-
|
Ah so they are! I can see why they would be useful there too. Yeah that seems like a good solution, thanks :) |
Beta Was this translation helpful? Give feedback.
-
|
Does Semantic Search detect intel iGPU support automatically or does a detector for OpenVINO need to be configured for ONIX support? If OpenVINO needs to be configured is there a way to specify which tasks to to which accelerator if I have a Coral for normal object detection already configured as well? |
Beta Was this translation helpful? Give feedback.
-
|
it will detect openvino automatically assuming you are using the main build and the GPU device is present. You can check this by using intel_gpu_top |
Beta Was this translation helpful? Give feedback.
-
|
So this is a win for the AI versus the native detection!
Could we maybe see in 0.16+ AI auto submitted, if toggled on of course, to Frigate+ when AI is right and the detector isn't? I know we have squirrel label options but I don't have that turned on at the moment but if it was a valid label just not added would be cool to train it properly. |
Beta Was this translation helpful? Give feedback.
-
|
A note for anyone running frigate on raspberry pi 5 with Hailo 8(L) chip: I randomly started getting bad inference times and found out I had to re-set the PCIe speed to 3 in rasp-config settings. I’m not sure why it reverted to 2 but this solved the inference time. |
Beta Was this translation helpful? Give feedback.
-
|
Out of curiosity, how many cameras are you running on your pi 5? Are you happy with the performance? |
Beta Was this translation helpful? Give feedback.
-
|
Right now, two, I’m new to frigate and home assistant and building it all from scratch. But with my Reolink POE doorbell, the raspberry pi 5 8gb running frigate in docker barely hits above 3% cpu usage, and the yolov8s model gets 30 ms inference time. The Hailo 8L chip barely goes above 3% either. The main thing I’ve noticed is that the pi 5 has not hardware encoding… :/ |
Beta Was this translation helpful? Give feedback.
-
How are you running Yolov8? I didn't think Frigate supported Yolov8 anymore? |
Beta Was this translation helpful? Give feedback.
-
|
I just tried 0.15.beta1 and it didnt seem to come up fully. I had 1/12 cameras show up on the main dashboard, and despite multiple restarts it didnt seem to come up fully. Nothing obvious in the logs. I shutdown and deleted my old DB, and restarted hoping that a fresh install would fix things but to no avail. I switched back to 0.14.1, swapped out the configs and DB for the backups I took and 0.14.1 came right back up, with all feeds showing within 30s or so. Something is not right with 0.15.0-b1 |
Beta Was this translation helpful? Give feedback.
-
|
Be sure to check out the release notes first. We'll need to see your logs, full config, and screenshots of the system metrics pages. Please open your own support discussion and we'd be happy to help further. |
Beta Was this translation helpful? Give feedback.
-
|
Understood - I did check out the release notes, but didnt see anything obvious other than perhaps ffmpeg v7? No shm messages in the log (Id already set a large chunk aside for that anyway). I dont have a GPU on the server running Frigate, but I do have a TPU and a lot of CPU/mem (dual Xeons) and that wasnt being taxed in anyway. |
Beta Was this translation helpful? Give feedback.
-
|
Anyone else notice that opening any specific life stream takes several seconds? I went to go look in Developer tools and what happens is there is activity for like 1s then a long pause of 2 seconds before something happens again. Opening a stream directly from go2rtc :1984/stream.html?src=doorbell is nearly instant so something seems to be happening when you click on the camera and then it seems to get stuck a little loading :5000/#doorbell |
Beta Was this translation helpful? Give feedback.
-
Ok but then why is frigate wildly all over the map 500ms to 8000ms slower to load the stream? |
Beta Was this translation helpful? Give feedback.
-
|
I am not seeing that, so I can't say for sure. Have you tried another browser? Does it happen with all your cameras or just one? |
Beta Was this translation helpful? Give feedback.
-
The slow load times are something I see on both my Windows Chrome browser and my iOS Safari browser it isn't consistant. It is also variable. I would say it is WORSE when there is a lot of tracked object activity. IE when I want to open a live camera from a notification or something. Right now no one is home and there is near zero activity on any cameras it is loading fairly fast only 500ms slower in frigate than direct in go2rtc. |
Beta Was this translation helpful? Give feedback.
-
|
Is it with all your cameras or just your doorbell camera? |
Beta Was this translation helpful? Give feedback.
-
I have 6 cameras I believe I have experienced slow loading on all of them at one time or another.. My Reolink WiFi Doorbell and my Reolink DUO2 just happen to be the most interesting cameras as they are outdoor cameras and I look at them the most and have notifications tied to them so when they don't load I get frustrated. |
Beta Was this translation helpful? Give feedback.
-
|
Hello, I am fiddling a bit with semantic search and genai - I have an (old) nvidia card available on my setup, so I started messing with ollama. My card is used also for stream encoding/deconding, and since it only has 4 GB of vram, I first started testing with I suspect there might be some timeout happening somewhere - I did not have the time to test the command line with vision, just chat, but I left the system going on its own for quite some time and none of the new events have a caption. I also monitor CPU usage when I request a caption, it pins to 100% for some time and then goes back to idle, so something is happening. Can anyone give any pointers as to how to debug/investigate this? Should I open an issue? Up to now it is not clear whether this is frigate timing out, ollama timing out, my machine shitting the bed or whatever, so I thought to ask here. Again thanks for this awesome piece of software. |
Beta Was this translation helpful? Give feedback.
-
|
You are correct - I am testing |
Beta Was this translation helpful? Give feedback.
-
|
You'll want to check out the docs for using Ollama: https://deploy-preview-13787--frigate-docs.netlify.app/configuration/genai/#ollama Parallel requests come with some caveats, and multi-modal parallel requests are currently not supported by Ollama. Depending on your hardware and the number of requests made to the Ollama API, these limitations may prevent Ollama from being an optimal solution for many users. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks @hawkeye217 - you are correct I most definitely need to fiddle with the options there. Perhaps requests are just dropped because they come while the one before is still being processed. I'll test it out and reply here with hopefully a solution, to help others who might find themselves in a similar predicament. |
Beta Was this translation helpful? Give feedback.
-
That's most likely the case. Other users have seem the same thing. It's a limitation of the llava model (and any Ollama multi-modal model) at this point. |
Beta Was this translation helpful? Give feedback.
-
|
FYI, we've updated the Ollama genai docs to include some Ollama environment variables you can set to improve the experience of using Ollama with Frigate. If you can, test it and let us know how it's working for you. |
Beta Was this translation helpful? Give feedback.
-
|
Not sure what it means yet, but just had 15b1 crash two days in a row with logs saying it's out of space for /dev/shm/log which obviously got changed a bit for this release. I did not get a notification SHM was too small which is why I bring it up here. I was setting it to 512mb of memory backed volume, upped it to 1024 for now. Is there a way I run the calculation manually to see if what was calculated really was not enough or if I missed the warning to make it bigger? |
Beta Was this translation helpful? Give feedback.
-
|
@NickM-27 when will beta 2 will be released? |
Beta Was this translation helpful? Give feedback.
-
Awesome, I will not attempt further digging unless I can be helpful to your work, let me know. Until then I'll just bounce the container when needed. Appreciate the note!! |
Beta Was this translation helpful? Give feedback.
-
There is no time frame available. We have a number of bugs still to work through. |
Beta Was this translation helpful? Give feedback.
-
|
if you want you can use |
Beta Was this translation helpful? Give feedback.
-
I can do that if it'd help, not a problem. |
Beta Was this translation helpful? Give feedback.
-
|
Are you able to add Groq to this Frigate version by chance? I tried using it with the OpenAI provider and I get the following: |
Beta Was this translation helpful? Give feedback.
-
|
There are a few open ai api compatible projects out there some acting as proxies for other services to enhance results, generalizing to allow custom URL etc would be useful for those. an example would be https://github.com/codelion/optillm Maybe call it openai_custom an and let the base URL be set and have a few other option parameters? |
Beta Was this translation helpful? Give feedback.
-
|
There is a PR up to make this possible, once merged you'll just need to set the |
Beta Was this translation helpful? Give feedback.
-
Thanks! Groq is actually different from Grok. Would it work the same since it's OpenAI compatible? |
Beta Was this translation helpful? Give feedback.
-
|
Sorry, I misread your original message. Any AI that fully supports the OpenAI API should work, though we'll only officially support the providers listed in the docs. |
Beta Was this translation helpful? Give feedback.
-
|
Hi all, Does anyone know why I can't playback my video recording from the Media saved recording? Thank you
|
Beta Was this translation helpful? Give feedback.
-
|
this is a known bug in HA 2024.11 |
Beta Was this translation helpful? Give feedback.
-
|
Great., thank you |
Beta Was this translation helpful? Give feedback.
-
|
Hello I've increased my dev/shm from 256 to 512. Frigate ran fine for a couple days and I got the message about running out of space. So I upped to 1024 and the message went away for a couple days until today.. how high should this be set? I have 5 cameras. |
Beta Was this translation helpful? Give feedback.
-
|
see #14653 (comment) |
Beta Was this translation helpful? Give feedback.
-
|
Thanks Nick I did see that. I'm just not very good with this stuff so changing an image I would probably break something in my setup. I was just curious if there was a quick fix or recommended number. I'll wait for beta 2 patiently. Thanks for all you guys do on this great project |
Beta Was this translation helpful? Give feedback.
-
|
So I am not sure exactly what is happing but on a LARGE number of my life cycle reviews the boxes and images don't come close. I was trouble shotting some false positives and found the boxes not clearly attached to anything. In others it looks like the boxes match the historic path of the object but the frames shown don't even include the object or it isn't even close. If I had to guess this got worse when I increased my detect FPS from 5 to 10 FPS to try and capture faster objects better however I think it has over all made tracking worse for all object? |
Beta Was this translation helpful? Give feedback.
-
|
Have you adjusted the annotation offset? |
Beta Was this translation helpful? Give feedback.
-
|
The object isn't even still visible to adjust to in some of the shots, and some of the objects are not fast moving objects. |
Beta Was this translation helpful? Give feedback.
-
|
I use nodered to mange notification frequency via HomeAssistant which includes following action to play the video clip related to the notification object. I am curious if similar link can be created so it will work like the native notification on 0.15 beta. I like how the native notification plays the recording inside the frigate explorer's review page |
Beta Was this translation helpful? Give feedback.
-
|
Looks like running out of SHM is still an issue, I am at a shm_size: "512mb" currently and things where running fine for a bit but I was doing lots of configs and restarts. left it for a day and sat down to find the UI hung and this going on again in the docker logs. |
Beta Was this translation helpful? Give feedback.
-
Is one of the new "features" of 0.15. This would be the second time I have had to increase it from it overflowing |
Beta Was this translation helpful? Give feedback.
-
|
We are aware of an issue that only affects some users and have a potential fix that is being tested |
Beta Was this translation helpful? Give feedback.
-
Is there a specific cause you need more confirmation on or is it more it just happens over time? |
Beta Was this translation helpful? Give feedback.
-
|
We believe we've found the bug and fixed it, feel free to pull the latest dev and test. |
Beta Was this translation helpful? Give feedback.
-
Is it a condition that can be forced? How far off are we from a beta 2? |
Beta Was this translation helpful? Give feedback.
-
|
Somebody else noticed that cars where accidently detected as Amazon or DPD? I've now had this on two occasions, on two different cameras. Looking in the debug screen gives a score of 96+%, the score the cars normally would have when being detected as actual cars. Restarting Frigate solves this; after the restart the cars were detected properly. |
Beta Was this translation helpful? Give feedback.
-
|
OK. I will enable alerts for cars on my camera's for debugging. Do I have to explicitly add all sub logos for this as well? Is there a way to delete multiple events in the explore section with a single go? Ctrl-A doesn't work, while it does so in the review section. |
Beta Was this translation helpful? Give feedback.
-
|
No, you can't multi-delete tracked objects in Explore. To be clear, deleting in Explore only removes the snapshot, any saved embeddings, and any associated object lifecycle entries. Recorded footage of tracked objects in History view will not be deleted. |
Beta Was this translation helpful? Give feedback.
-
|
This is exactly what I want. Will this do the job? snapshots:
retain:
objects:
car: 2 |
Beta Was this translation helpful? Give feedback.
-
I noticed the same, in my case I had to go through several of the reviews to see it was delivery trucks driving past parked cars and the logo migrating to the parked car. |
Beta Was this translation helpful? Give feedback.
-
|
@NickM-27 saw your last PR. No action needed, unless it happens again in beta 2? |
Beta Was this translation helpful? Give feedback.
-
|
I'm getting a lot of hanging car tracking.. like 45min to hour long clips of parked cars in driveway. A car will drive by which will trigger the detection but then it focuses on another parked car in the background and stays there. How could I eliminate this? |
Beta Was this translation helpful? Give feedback.
-
|
In the tracked object page in the car section. Here is 25 minute clip lifecycle. In this case the first car detected was a parked car in a driveway across the street. It never moved once during entire clip. At 8:31:12 another vehicle entered the zone and then left and thats what ended the clip. |
Beta Was this translation helpful? Give feedback.
-
|
you are looking at an object, objects are tracked as long as they are detected so that is working as expected. If the review items are not running that long |
Beta Was this translation helpful? Give feedback.
-
|
No I don't have cars in the road set for review.. Does tracking cause a big strain on the gpu? The reason I saw this was bottom right of my frigate it said driveway camera using high fffmpg gpu usage, so when I looked at the object page one of these cars were being tracked.. |
Beta Was this translation helpful? Give feedback.
-
|
I think you are mixing things up. The bottom right will only have a message that ffmpeg is using high CPU. And, that has nothing to do with tracking |
Beta Was this translation helpful? Give feedback.
-
|
It's quite probable I'm mixing things up lol. Well thanks for taking the time to answer my questions. I'll leave it alone if all is working as intended. |
Beta Was this translation helpful? Give feedback.
-
|
What's the normal cadence (if there is one) for beta releases? |
Beta Was this translation helpful? Give feedback.
-
|
there isn't one, just when they're ready |
Beta Was this translation helpful? Give feedback.
and others
-
Beta Documentation: https://deploy-preview-13787--frigate-docs.netlify.app/
Images
ghcr.io/blakeblackshear/frigate:0.15.0-beta1ghcr.io/blakeblackshear/frigate:0.15.0-beta1-standard-arm64ghcr.io/blakeblackshear/frigate:0.15.0-beta1-tensorrtghcr.io/blakeblackshear/frigate:0.15.0-beta1-tensorrt-jp4ghcr.io/blakeblackshear/frigate:0.15.0-beta1-tensorrt-jp5ghcr.io/blakeblackshear/frigate:0.15.0-beta1-rkghcr.io/blakeblackshear/frigate:0.15.0-beta1-rocmghcr.io/blakeblackshear/frigate:0.15.0-beta1-h8lMajor Changes for 0.15.0
Breaking Changes
There are several breaking changes in this release, Frigate will attempt to update the configuration automatically. In some cases manual changes may be required. It is always recommended to back up your current config and database before upgrading:
frigate.dbfileshm_sizeis too low then a warning will be printed in the log stating that it needs to be increased.recordconfig has been refactored to allow for direct control of how longalertsanddetectionsare retained. These values will be automatically populated from your current config, but you may want to adjust the values that are set after updating. See the updated docs here.hwaccelpreset they are using (preset-vaapimay need to now bepreset-intel-qsv-h264orpreset-intel-qsv-h265).ffmpegis no longer part of$PATH. In most cases this is handled automatically, but go2rtcexecstreams will need to add the full path for ffmpeg which in most cases will be/usr/lib/ffmpeg/7.0/bin/ffmpegExplore
The new Explore pane in Frigate 0.15 makes it easy to explore every object tracked by Frigate. It offers a variety of filters and supports keyword and phrase-based text search, searching for similar images, and searching through descriptive text generated by AI models.
AI/ML Search
Frigate 0.15 introduces two powerful search features: Semantic Search and GenAI Search. Semantic Search can be enabled on its own, while GenAI Search works in addition to Semantic Search.
Semantic Search
Semantic Search uses a CLIP model to generate embeddings (numerical representations of images) for the thumbnails of your tracked objects, enabling searches based on text descriptions or visual similarity. This is all done locally.
For instance, if Frigate detects and tracks a car, you can use similarity search to see other instances where Frigate detected and tracked that same car. You can also quickly search your tracked objects using an "image caption" approach. Searching for "red car driving on a residential street" or "person in a blue shirt walking on the sidewalk at dawn" or even "a person wearing a black t-shirt with the word 'SPORT' on it" will produce some stunning results.
Semantic Search works by running an AI model locally on your system. Small or underpowered systems like a Raspberry Pi will not run Semantic Search reliably or at all. A dedicated GPU and 16GB of RAM is recommended for best performance.
See the Semantic Search docs for system requirements, setup instructions, and usage tips.
Generative AI
GenAI Search employs generative AI models to create descriptive text for the thumbnails of your tracked objects, which are stored in the Frigate database to enhance future searches. Supported providers include Google Gemini, Ollama, and OpenAI, so you can choose whether you want to send data to the cloud or use a locally hosted provider.
See the GenAI docs for setup instructions and use case suggestions.
Improved Tools for Debugging
Review Item Details Pane
A new Review Item Details pane can be viewed by clicking / tapping on the gray chip on a review item in the Review pane. This shows more information about the review item as well as thumbnails or snapshots for individual objects (if enabled). The pane also provides links to share the review item, download it, submit images to Frigate+, view object lifecycles, and more.
Object Lifecycle Pane
The Recordings Timeline from Frigate 0.13 has been improved upon and returns to 0.15 as the
Object Lifecycle, viewable in the Review Details pane as well as the new Explore page. The new pane shows the significant moments during the object's lifecycle: when it was first seen, when it entered a zone, became stationary, etc. It also provides information about the object's area and size ratio to assist in configuring Frigate to tune out false positives.Native Notifications
Frigate now supports notifications using the WebPush protocol. This allows Frigate to deliver notifications to devices that have registered to receive notifications in the Frigate settings, delivering them in a timely and secure manner. Currently, notifications will be delivered for all review items marked as alerts. More options for native notifications will be supported in the future.
See the notifications docs.
New Object Detectors
ONNX
ONNX is an open model standard which allows for a single model format that can run on different types of GPUs. The
default,tensorrt, androcmFrigate build variants include GPU support for efficient object detection via ONNX models, simplifying configuration and support more models. There are no default included ONNX models for object detection.AMD MiGraphX
Support has been added for AMD GPUs via ROCm and MiGraphX. Currently there is no default included model for this detector.
Hailo-8
Support has been added for the Hailo8 and Hailo-8L hardware for object detection on both arm64 and amd64 platforms.
Other UI Changes
Other Backend Changes
This discussion was created from the release 0.15.0 Beta 1.
Beta Was this translation helpful? Give feedback.
All reactions