Blue Iris

Yup, thank goodness for good back ups

The screen had gone black, but I thought I would leave it the night to see if BI5 and DS could work like this. I don't like the idea of not knowing what is happening on a pc, but I could tell things like Homeseer were actually working. Later I tried to turn off the lounge AV equipment (IR OFF on remote control) to find the VH UsbUirt was not working. Server OFF and go to bed !

I have booted it up this morning, and restored it to before DeepStack was installed. I think my DS adventure ends here for the moment

Stuff wot I learned:
1. You don't need sub streams if your cameras can do Onvif motion detection. The Dahua 5231 is catching the same motion as BI5 does, but ignores some of the moving shadows. This must be reducing cpu loading.
2. DeepStack needs the BI5 motion images, else it won't do anything. Onvif triggers won't work.
3. DeepStack and BI5 together can black out your screen. Make sure you do regular back ups before installing new software.
4. The screen didn't black out until I cloned an Onvif triggered camera and made it a DeepStack camera.

And all of this didn't affect my primary BI5 installation running on the other server

Update...

I was able to confirm DS was receiving the alert images. One of the DS processes would increase for several seconds.
But, nothing returned to BI!

Well, being that DS (on my older desktop) will use up to 45% cpu, I decided to scrap DS. Even if I could get the response into BI, DS uses too much cpu to make it usable on my current setup. I may retry when I build my new box.

One interesting note... a few hours after I uninstalled DS and disabled it in BI, all my content went blank in Windows 10 and all programs. I first noticed this in BI.. nothing in clip list and timeline, Windows content blank...
Rebooting Windows resolved this, thankfully.

Glad to hear you got it all working. You are right that DS on Windows needs a decent amount of processing power.

My experience with DS has been pretty good over all, but I have been using DS with AITools (by GentlePumpkin from IPCamTalk) for a while so I had a good deal of experience with getting DS running at the command prompt both in Windows and the Docker container.
As a note, you don't need PowerShell to run DS on Windows if you want to start it manually. You can just create a shortcut on your desktop, or go to the standard command prompt and run DS from it's folder just including the options.

I did try it without substreams for a while, but I found the accuracy not greatly improved just for object detection. DS is pretty good at detecting objects from low res images. If you do pass high res images to DS and you are using the CPU version it tends to by quite power hungry. I have several cameras and a bit of a busy environment and a single DS server just didn't keep up well with high res images so I grab the sub-stream and that reduced the delays to a tolerable level. I don't use it for facial recognition and object detection is virtually identical in accuracy with the low res image.

I ended up dumping DS into a Virtual Machine and ran the docker image though as the Docker image running under linux uses much less CPU time and also keeps it separate from Windows so if DS does have a hissy fit, the rest of BI keeps running. My Server also runs Home Assistant which is my home automation solution. I also use DeConz for Zigbee. Never had an issue with DS killing DeConz.

Obviously this is my experience with DS and BI and every experience is unique.

As a note, I am running 4 HikVision DS-2CE56H5T-IT3-B and 3 Reolink cameras, 2 RLC-520 and 1 RLC-820. The Reolinks will be going very soon as their night vision is disgusting. BI Server is an Intel Core i7 7700 with 32GB RAM and 16TB storage (1tb SSD for scratch/cache).

Closing this tread, and starting a new one.