#help setting up frigate to constantly record and also getting smooth playback of streams in HA
1 messages · Page 1 of 1 (latest)
the default cameras that are added by the frigate integration has maybe a frame rate at 1. when clicking the camera it updates maybe once or twice per second
when viewing directly in frigate the frame rate is better, even though it's not perfect.
viewing in go2rtc is perfect
on iphone it's even worse
that is the normal behavior when go2rtc is not setup correctly, so a snapshot that updates once per second is shown
how are you displaying the streams in Home Assistant? what card?
i've tried both the frigate card and the advanced camera card
i'm running frigate lxc container and that has its own go2rtc. and i'm using that
what other stuff do i need to do to configure go2rtc correctly?
that is. the default dashboard is automatically populated with some camera entities. so i added the advanced camera card and that works better, but only for android and pc users. for iphone it's still terrible
You'll want to make sure and use the go2rtc provider in the advance camera card
Make sure the go2rtc stream names match the camera names
Is it the same when using frigate directly on iPhone? That's what I use for frigate (PWA) on iPhone and it works without any issues
another question. looking at frigate system metrics i see under storage that there seems to be a storage limit of 200 GB? what's what all about? and how can i changed that?
78.68 GiB / 200.89 GiB 39%
but there's still 80 TB available
and i want to store everything
200GB is the size that Frigate sees on the drive configured for recording
so either the recording directory is not mounted on the correct storage, or there is some virtual filesystem with a storage limit on it (this is common in Proxmox for example)
Frigate itself imposes no such limit
i have bind mounted a volume, yes
ok, so frigate won't actually start deleting recordings?
because that would make me a sad panda 🙂
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/pve-vm--104--disk--0 20G 16G 3.2G 83% /
192.168.227.13:/tank/nvr/frigate 70T 1.2T 68T 2% /media/frigate
i'm not sure where it gets 200GB from, though?
oh, it's actually increased to 230-ish
oh
that is just telling you the total of how much storage has been used
recordings are using 320GB out of the available 80TB
you are also on an older version of Frigate, 0.15 has a re-organized storage graph
i'm running frigate in an lxc container on proxmox
so i'm using the version that's bundled there
is it possible to train my own AI models? i'm seeing some information about that, but i'm not getting a conclusive answer on the matter
it is certainly possible, generally training an object detection model is quite complicated if you are going to do it by hand. It requires tens of thousands of images.
Frigate+ is a service which offers this for a fee, allowing users to annotate images and get an object detection model that is fine-tuned on their cameras https://frigate.video/plus/
yeah, but i want to track/classify fish
i see that classification models isn't available so far..?
i have trained a model before for fish
so i was just wondering if it's possible to use this in frigate in a way
I have implemented support for classification models in 0.17, but that version is under early development
first and foremost i want to just count the number of fish, but the next step is being able to classify them. we want to know the species
i'd love to use frigate+ for this, but it needs to support the fish label 🙂
but i really suspect frigate maybe isn't the right tool for this?
Yeah if detecting fish is the goal, it could be accomplished in a simpler manor using software that isn't primarily an NVR
do you have any suggestions on how to do that? because i don't want to do object detection on every single frame. i only want to do this on motion, for instance
it seems like most of the implementations rely on writing your own code. which i'm not against, per se, but if there's already a framework available out there..