@SgtBatten OMG, THANK YOU!!! I was totally blind 
it is soooo easy
1 Like
Does anyone know why I canāt download clips or snapshots from Frigate in the iOS companion app but I can with Safari?
Does anyone know about using different models? Iām pretty happy with the default Frigate models but thereās two different things Iād like to do:
-
add a āchickenā model as currently Frigate canāt reliably detect them. I was hoping to be able to use Frigate+ to train images of my actual chickens, not sure when that is going to be available though. I see the Coral sample models include a āhenā object which I assume would work okā¦ is there anyway to extract this and add it to the Frigate models? I assume Frigate uses a sub-set of the Coral example models, no idea if itās possible to create my own sub-set that includes āhenā though? I assume I could just use the entire Coral model set, itās over 1000 objects though the vast majority of which I wonāt need, are there any performance implications of doing this?
-
Iād love to be able to classify the bird species that visit my bird tableā¦ although the Coral example models include many bird species, I see thereās a separate ābirdā model with 900 bird species. Is it possible to hload 2 different models into Frigate? I would probably only want to run the bird model against a single camera. Would it be worth running a 2nd instance of Frigate just to handle bird classification? Is that even possible? Have a reasonable system & multiple Corals available. Any other ideas appreciated!
EDIT: could use DOODS I guess for the bird classification with a separate model, would be good to be able to use Frigateās events though to avoid counting a bird more than once!
Hi folks,
I read this on dockerhub
# NOTICE: This repository will no longer being updated beginning with version 0.12.0.
Latest images can be found on ghcr instead: https://github.com/blakeblackshear/frigate/pkgs/container/frigate
how I should replace my docker-compose.yml file to consume it?
image: blakeblackshear/frigate:stable-aarch64`
version: '3.9'
services:
frigate_camera:
privileged: true
container_name: frigate_camera_tpu
restart: unless-stopped
image: blakeblackshear/frigate:stable-aarch64
I donāt use docker-compose.yaml but itās probably just adding the ghcr.io
this is what I use
-d ghcr.io/blakeblackshear/frigate:0.12.0-beta7
1 Like
Iām unable to start the Frigate Proxy add-on within HA
Frigate is running OK in docker on another system
this is what the log says:
(the server address is definitely correct as when I go to that address get the Frigate page & can see camera stream etc)
Any ideas?
EDIT: never mind, had problems with other (although not all!) add-ons too, reboot of host (rather than just HA restart) sorted it, will leave this post up in case anyone else has the same issue!
you are on an old version of frigate. For now (to stay on stable) you should be using blakeblackshear/frigate:stable-aarch64 and once 0.12 goes stable you can use ghcr.io/blakeblackshear/frigate:stable or as another user said if you want to get onto the beta you can use ghcr.io/blakeblackshear/frigate:0.12.0-beta8 right now
2 Likes
Hey guyz,
Please somebody help/explain.
I have HAOS on Proxmox with TPU coral and GPU nvidia pasthrough. Coral is working perfectly but gpu for decoding is not working, its always errors like:
[2023-02-19 23:06:22] frigate.video ERROR : CAM6: Unable to read frames from ffmpeg process.
[2023-02-19 23:06:22] frigate.video ERROR : CAM6: ffmpeg process is not running. exiting capture
hwaccel_args:
- -hwaccel
- qsv
- -qsv_device
- /dev/dri/renderD128
input_args: -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 5000000 -use_wallclock_as_timestamps 1# output_args:
detect: -f rawvideo -pix_fmt yuv420p
I donāt know how to let it accelerate, i think i tryed everything. I canāt use CPU gpu as itās a Xeon server with ILO
Protection mode is off (addon version with full access)
Please advise
if your CPU doesnāt support it, then qsv isnāt going to workā¦ youāll need to check if your cpu has support for VAAPI
1 Like
If you have an nvidia GPU then qsv is incorrect as that is for intel. Also, nvidia is not supported in the addons
1 Like
MNOOOOOOOO 
So i need to make a container for frigate,seperate from home assitant?
If you want to use an nvidia GPU, yes
1 Like
Inference time is 6.87ms means it is being seen and used. Whats the issue you having?
Ah ok. I assume itās only used for detection, not processing of the stream so my normal CPU is still being used quite a bit?
Youāre using your CPU for stream decoding which will use a lot. If you have an integrated GPU should use that.
Yes
TPU is only used for detection. All other tasks are CPU.
GPU can be used in detection but not needed if you have TPU. Not certain how GPU is handled for other tasks. I would presume it gets used as needed and will reduce cou load. You will need to add it to docker container so the container knows its available for use.
Friends, please tell me where I made a mistake? Does not delete the video after 2 days.
`
camera_4: # <------ Name the camera
mqtt:
crop: True
height: 500
rtmp:
enabled: false
ffmpeg:
inputs:
- path: rtsp://admin:[email protected]:554/2 # <----- Update for your camera
roles:
- detect
- record
- clips
detect:
width: 1280 # <---- update for your camera's resolution
height: 720 # <---- update for your camera's resolution
fps: 14
objects:
track:
- person
- car
- bird
- cat
- dog
record:
enabled: True
events:
retain:
default: 2
If I change
events:
retain:
default: 2
on
events:
retain:
day: 2
`
Frigate is not working. Writes an error
I have this
record:
enabled: True
retain:
days: 15
mode: motion
events:
pre_capture: 15
post_capture: 15
objects:
- person
retain:
default: 10
mode: motion
1 Like