OS or user issues

#3

I’ve been running them in containers, but now one of them doesn’t start. Anyway I need to start from scratch.

Can someone tell me, should I update the system or not?

#4

If You are using radxa-os then update it using rconfig (there are some issues with packages when it’s done via apt). Since Armbian support NPU it may worth to try with it,

here it’s documented for opi5+ and coral, You may find this useful:
https://codingfield.com/blog/2024-01/install-frigate-in-proxmox-on-orangepi5plus/

1 Like
#5

Or save yourself some trouble and use Armbian or Joshua’s Ubuntu

1 Like
#6

Thanks I was sitting on radxa-os and I was using rconfig, but it was causing so much issues… I didn’t know that others support NPU and it is what I need, I will try armbian then! Maybe will be easier, as radxa-os gives me headaches.

The guide you provided is from Jan 2024, not sure why they went with coral TPU as RK3588 got nice NPU, anyway I managed to sort NPU yesterday but after docker restart it started crashing etc. Also Frigate does come with nice image for rock chips:

the one that ends with - rk is optimised for this NPU

1 Like
#7

Remember that ARM is not x86. At this point, don’t expect mainline ffmpeg to support the Rockchip VPU. Your options are ffmpeg-rockchip or just using jellyfin’s ffmpeg. Just install it and alias/link as ffmpeg

#8

Check the download options and pick your poison… likely anything but radxa official is better supported… Joshua stopped supporting / armbian is always there and if you like arch linux / Kwankius installer is the best option

#9

Thanks! I went with armbian as Joshua’s ubuntu us no longer updated.

Anyway, looks like everything(so far) is working withotu any issues on the armbian, I am surprised that radxa’s debian is so bad… But at least I got my system working and can start doing the right stuff.

#10

Its like this: u dont buy an lenovo laptop asus acer hp and expect all of them to give u their own version of windows… Radxa provides and you tweak it… but everyone uses armbian thats also the reason why they are more relaxed… cause at the end of the day the Pull Requests goes to armbian.

#11

Of course I do not expect them, moreover I do not expect them to maintain the drivers, however, if radxa is supplying us os, at least they should make sure that we are not loosing Ethernet drivers after restart and our fan is not doing 120% in idle :smiley:

yeah armbian is the way to go from now for me :slight_smile: I was disappointed initially but now I am getting happy with the device. Just need to understood how to move the system to the nvme :smiley:

3 Likes
#12

Armbian provides script to do that, armbian-install or something like that, it will do most job for You, updating SPI to right image, move system etc. If Your nvme is compatible enough and there is some power for it then it should just boot.

If You find armbian useful You can post guide for others what is up to date in 2025 with frigate and rock 5b. I hope You will get all needed stuff to work as expected. Armbian has own issues and don’t be surprised if something will be broken just on update, but for sure it’s solid base to get somewhere :smiley:

1 Like
#13

BTW: opensource Rockchip NPU driver is expected in linux 6.18

This will make at least few things easier :slight_smile:

#14

Joloxx9, Many thanks for keeping this post updated with your progress; it’s a huge help for the rest of us newbies.

#16

What update do you expect? I’ve said that I’ve movwd to armbian as it is working out of the box.

#17

I run the Rock 5B with Frigate, here are my install instructions.

Rock 5B - Frigate Setup

Install RadaxOS debian 12 bookworm image.

OS Update

Don’t do apt upgrade to update the debian system, instead to update do;

rsetup

Then from the menu select System->System Update.

First Changes

Disable Desktop booting, so it boots to CLI only.

sudo systemctl set-default multi-user.target
reboot

Install common tools

apt install net-tools build-essential dnsutils

Install Docker

Install dependencies

sudo apt install apt-transport-https ca-certificates curl gnupg lsb-release

Add Dockers GPG Key

curl -fsSL https://download.docker.com/linux/debian/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg

Add docker APT repository

echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/debian $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

Update APT repo information

sudo apt update

Install docker

sudo apt install docker-ce docker-ce-cli containerd.io docker-compose-plugin docker-buildx-plugin

Add user to docker for permission access

sudo usermod -aG docker rock

Logout and log back in SSH session.

Verify installation as user

docker run hello-world

Install Frigate

Create a root project directory for frigate, I did this under the rock linux user.

mkdir ~/frigate
cd ~/frigate

Create frigate directory structure

mkdir storage config && touch docker-compose.yml

Edit the docker-compose.yml and set contents;

services:
  frigate:
    container_name: frigate
    restart: unless-stopped
    stop_grace_period: 30s
    image: ghcr.io/blakeblackshear/frigate:stable-rk
    shm_size: "1024mb"
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - ./config:/config
      - ./storage:/media/frigate
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "8971:8971"
      - "8554:8554" # RTSP feeds
    security_opt:
      - apparmor=unconfined
      - systempaths=unconfined

    devices:    
      - /dev/dri
      - /dev/dma_heap
      - /dev/rga
      - /dev/mpp_service

Start frigate with;

docker compose up -d

Get admin password generated from logs;

docker logs frigate

Go to web interface on port 8971, ie: https://<rock5b ip address>:8971

Configure Frigate

Edit the frigate configuration in the Web UI and setup the cameras and object detectors.

Used config of;

mqtt:
  enabled: false


ffmpeg:
  hwaccel_args: preset-rk-h264

detectors:
  rknn1:
    type: rknn
    num_cores: 3
  rknn2:
    type: rknn
    num_cores: 3
  rknn3:
    type: rknn
    num_cores: 3

model:
  # name of model (will be automatically downloaded) or path to your own .rknn model file
  # possible values are:
  # - deci-fp16-yolonas_s
  # - deci-fp16-yolonas_m
  # - deci-fp16-yolonas_l
  # - /config/model_cache/your_custom_model.rknn
  path: deci-fp16-yolonas_s
  # width and height of detection frames
  width: 320
  height: 320
  # pixel format of detection frame
  # default value is rgb but yolo models usually use bgr format
  input_pixel_format: bgr
  # shape of detection frame
  input_tensor: nhwc
  # needs to be adjusted to model, see below
  labelmap_path: /labelmap.txt

#
# configure your camera streams here
#          
          
version: 0.15-1

Start at Bootup

Create a systemd service so frigate starts on boot.

Create file /etc/systemd/system/frigate.service with contents;

[Unit]
Description=Frigate NVR
Requires=docker.service
After=docker.service network-online.target
Wants=network-online.target

[Service]
Type=oneshot
RemainAfterExit=yes
WorkingDirectory=/home/rock/frigate
User=rock
Group=rock
# If you use Docker Compose v2 plugin:
ExecStart=/usr/bin/docker compose up -d
ExecStop=/usr/bin/docker compose down
TimeoutStartSec=0

[Install]
WantedBy=multi-user.target

Enable;

sudo systemctl daemon-reload
sudo systemctl enable frigate.service

Reboot and test.

1 Like
#18

Thanks for those,
have You measured power consumption and load of those? How much cameras do You have?

#19

I have 3x 2k cameras, 3 streams 2k + 3x low quality streams for object detection, everything with HW acceleration and with ffmpeg codec h264, NPU usage is roughly 20-30%, CPU ffmpeg on the other hand is 5%.
I will paste my config later on.

1 Like
#20

I only run two 4k cameras with CPU load around 18%. I have never measured power consumption so not sure about that.

#21

If you install ffmpeg and switch to HW acceleration then you would have lower cpu usage.

#22

I guess that frigate should work on smaller previews? What kind of stream do both of You have now?
Power consumption should be important thing to consider with 24/7 stuff, bsp kernel idles at 1.5W, edge at about 4W, but this may be not that important when cpu has some constant workload. For sure hardware acceleration/npu should help with those, especially face detection. I’m curious how much this will take power for those tasks.

#23

I am not afraid about power consumption, my unraid server eats too much anyway.

Thisnis my config:

3x 2k live previewnand low res streams dor frigate AI detection which is only set to test that it is working, I nead reconfigure it anyway