Thoughts on running CodeProject.AI or Deepstack on remote server or PC for BI

[dbrannon79]

Good morning everyone, I had the thought occur to me about the possibility of running the AI inside a docker container of a remote server within the same network to alleviate some of the load on my current PC running BI.

I have TrueNAS Scale running on a Dell R720 server that really doesn't do much other than house a lot of nas storage space for my home network. I have a couple of docker apps running on it like Plex and Adguard for the house, but I was thinking if I could utilize the servers processing power for AI to be used for the BI pc which is in the same server rack as the dell r720.

I looked into the docker library availability and see that both CodeProject.AI and DeepStack are both available in docker containers. I could possibly set this up and have a dedicated lan port from the dell server run to the BI pc assigning static IP addresses outside of my main network if needed just to give it's own tunnel and not bog down my network that supplies the NAS and internet throughout the house, but I think it would run fine over the main network.

has anyone ever attempted to separate BI and the AI on different machines? would there be any benefit to running AI on a different machine with available CPU horsepower to spare.

 

[jjj666]

I currently use CodeProjectAI (as a windows VM running under Proxmox) on a R710 with a dedicated p4000 Nvidia card installed for AI processing. I have 6 NVRs feeding it captures for AI processing and seems to be running good. Much less false alarms being populated. I use synology SS instead of BI though. Depending on the number of cameras, im sure you could get away with the CPU processing AI, unless you can utilize a GPU in the codeprojectai docker. Im still new to Codeproject, still trying to figure out how to schedule different models to use for different times (day/night), but overall its been fine.

 

[Vettester]

has anyone ever attempted to separate BI and the AI on different machines?

In an attempt to load balance I'm running two instances of CPAI. The first instance runs on my BI server (i7-4790) while the second instance runs on my HA server (i7-6700T). The instance running on my BI server is processing all of my LPR images which is somewhere in range of 900 - 1200 cars per day. The HA machine is processing images from 14 cameras along with 3 clones and everything is running well.

 

[dbrannon79]

Well I gave it a shot. I installed CodeProject.AI as a docker on my Dell R720 ( running dual Xeon E-2620 cpus) gave the container it's own IP and the respective port number. went into BI on the dedicated PC and changed to local host ip to the ip where the docker was running unchecked "start AI with BI" and saved changes. and it's working HAHA.

the funny thing about this is that the docker image is running version 2.1.8 I could only get 1.6.8 to work running on the BI pc. strange, but I'm not complaining. the highest CPU usage I have seen on the server is 80% but that was when all 6 cameras went on alerts at the same time.


now I am seeing something kinda strange with the CodeProject.AI that I left running on the BI pc. it's not set to handle BI anymore but seems to be picking up were docker maybe misses ?? the CPU usage is staying low on the BI pc as appose to it running at full tilt before, but I'm not sure if I should leave the AI running on the BI pc. if BI is using it as a fall back I will leave it.

any ideas?