Deepstack as Docker

[stev067]

Is anyone running Deepstack as a docker container? I wanted to try it out to see if I could gain any speed improvements by running it as a docker on my Unraid server, as opposed to how I have it now, installed on the same Windows 10 VM that Blue Iris runs on. From initial tests, it's taking a lot longer to process images. Just wondered if anyone has set this up, and if they prefer it or not.
PS I'm running the CPU version and intend to keep it that way for the time being.

 

[whoami ™]

Using AITool v.2.0 from October 2020 I've run DS CPU version on Docker Desktop on the same Windows 10 machine as BI, average time I was getting was around 300ms. Then I tried running windows version of DS CPU on the same machine and got the same 300ms average. Then I bought a Nvidia Quadro P400 and tried running the Windows GPU version of DS on the same windows machine and was getting the same 300ms average.

Then I updated to the latest version of AITool 2.0.1206 running the Windows GPU version of DS and am getting around 100ms average. AITool was updated at some point in time to be able to run multiple instances of DS for parallel processing. Running two instances of DS I was getting an average of 35ms with 8 cams. But I was having issues with AITool using a lot more memory when running multiple instances of DS, I had to add another 12gb of RAM to my system to keep it at a reasonable level. I've found that its better just running one instance with an average of 100ms, that's only 1/10th of a second.

I have no experance running DS directly from Blue Iris. My updates expired a year ago and I don't plan on buying another year of updates until they get all the kinks worked out and running DS in BI works better than AITool.