New Script Suggestion: CodeProject.AI

[xijio]

https://hub.docker.com/r/codeproject/ai-server

https://www.codeproject.com/Articles/5322557/CodeProject-AI-Server-AI-the-easy-way

This is an open source AI object detection engine, a re-write of the abandoned deepstack. It's frequently used in companion to Blue Iris as the server for smart object detection for home security footage. I currently have it installed in a docker lxc, but would be awesome to have it's own script.

[lightmaster]

Would love to see this. I ended up creating a Docker LXC and installing CodeProject.AI Server in that instance of Docker. I'm sure there's some loss of performance due to the nesting there, but better than setting it up in a VM. Would be awesome to get a script that sets CodeProject.AI up directly inside a LXC

[Yonny24]

Did you get CPAI working as an lxc on proxmox? I've tried several times, it's running but not working 100%

Testing any image just gets queued and not processed.
Any ideas?

Are you running it on a debian or ubuntu lxc? Unprivileged with host network?

Thanks

[lightmaster]

Sorry it took me so long to respond, I never get any emails from github about responses or anything.

So I have CPAI running as a Debian 12 LXC. I used the Docker LXC template, selecting Debian 12 and for it to be Privileged. I gave it 8 CPU cores, 4GB RAM, and 20GB SSD. I also have Nesting enabled (pretty sure template did that). Then I used Portainer to create a new Stack, using the below docker-compose.yml. I specify the tag number so that I know it won't ever pull an update unless I want it updated (got burned twice by bad updates that just didn't work).

I also have custom models installed to `/opt/codeproject/custom-models/YOLOv5-6.2, downloaded from https://github.com/MikeLud/CodeProject.AI-Custom-IPcam-Models/tree/main/PT%20Models

In my router, I have a DHCP reservation setup for the LXC as codeproject-ai, so that I can run ping codeproject-ai from Windows Powershell on my BI server (or you could just use the IP instead of the hostname). Then I have BI setup to use the AI server on codeproject-ai and port 32168 and a timeout of 15s. Make sure to uncheck "Auto start/stop with BI". If you want to use custom models, check that option. I have my cameras set to search for person,bicycle,vehicle,car,motorcycle,bus,truck,cat,dog,deer to confirm in Camera > Ai Settings

version: "3.9"

services:
  CodeProject.AI-Server:
    image: codeproject/ai-server:cpu-2.5.4
    container_name: CodeProject.AI-Server
    volumes:
      - /opt/codeproject/custom-models/YOLOv5-6.2:/app/preinstalled-modules/ObjectDetectionYOLOv5-6.2/custom-models
      - /etc/localtime:/etc/localtime:ro
    environment:
      - TZ=America/New_York
    ports:
      - 32168:32168/tcp
    restart: unless-stopped

Then, when the Docker LXC is running and CPAI is running inside Portainer, you can go to http://IP:32168 in a browser and should see the interface, with the count ticking up on YOLOv5 6.2 whenever BI sends an image to it.

[xijio]

Also working for me. I used the docker lxc and here is my docker-compose.yml:

version: "3.9"

services:
  CodeProjectAI:
    image: codeproject/ai-server:cuda11_7
    container_name: codeproject-ai-server-cuda
    hostname: codeproject-ai-server
    restart: unless-stopped
    ports:
      - "32168:32168"
    environment:
      - TZ=America/Los_Angeles
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

I used this guide for getting the nvidia gpu inside: https://jocke.no/2022/02/23/plex-gpu-transcoding-in-docker-on-lxc-on-proxmox/

[Mrdindon]

Yeh that would be great... CPAI is moving fast and have it scripted would be nice to test newer versions.
Currently running it on windows vm with Coral usb accelerator passthrough. (For BlueIris :) )

[Mrdindon]

Any chance seeing this automated with your deployments scripts ? :)

[SleepinDevil]

Inline with the discussion in here, I've been playing with some LXC containers to get CodeProject.AI working directly on an LXC container running Ubuntu instead of the 2nd layer through Docker as well. No success yet, but do want to chime in to the sentiment here that a script for it would be much appreciated. My knowledge of Linux is very limited 😅 so whatever I manage will be a hackjob and likely going to be a privileged container too as I am trying to get an AMD GPU and USB/PCIE Corals working too. A privileged LXC which aficionados of Proxmox very strongly steer people away from...