BEWARE: Issue with Deepstack 2022.01.01 and BlueIris

MikeLud1

IPCT Contributor
Joined
Apr 5, 2017
Messages
2,141
Reaction score
4,118
Location
Brooklyn, NY
I'm having the same issues- Windows, GPU version, no objects are returned except for the following:

{
"api":"general",
"found":{
"success":true,
"predictions":[
]
,
"duration":0}
}
]

Running latest BlueIris 5.5.5.1 which I think has 2022.01.1 support?

I'm still on Cuda 10.1 as per the DeepStack instructions and associated cuDNN version.

Any ideas?
I think you need to use cuda 11.6 and cudnn 8.3.2. That is what I am using with my GPU.
 

Lucas Tam

n3wb
Joined
Apr 12, 2017
Messages
27
Reaction score
5
I think you need to use cuda 11.6 and cudnn 8.3.2. That is what I am using with my GPU.
Thanks - the documentation on DeepStack's website still refer to Cuda 10.1. Release notes are also "To be Updated" - looks like this may have prematurely released this version? :D

I'll give 11.6 a try, do you know if you can run multiple CUDAs in parallel?
 

austwhite

Getting the hang of it
Joined
Jun 7, 2020
Messages
93
Reaction score
92
Location
USA
It does seem DeepStack 2022.1 has had some big changes. Firstly, it uses almost twice the disk space as the 2021.9 version.
The newest Blue Iris versions work with it through the internal integration, but older versions of Blue Iris will need 2021.9.
If you use the old third party integration or older Blue Iris version, get Deepstack 2021.9 and don't update to 2022.1

Edit: My note is about CPU version, I have not tested the GPU version as I currently don't have a spare NVIDIA card to test it with.
 

Skispcs

Getting the hang of it
Joined
Jun 10, 2014
Messages
104
Reaction score
30
For what it is worth, in June I upgraded Deepstack GPU to the newest version and the same thing happened.
No Objects detected.
Rolled back to 21-9 and Blue Iris detections started working again.
 

iwanttosee

Pulling my weight
Joined
Dec 27, 2020
Messages
203
Reaction score
186
Location
US
I'm using Proxmox so I can monitor ram usage easily. I'm using Windows 10 VM, P620 GPU bypass, CUDA10, DeepStack-Installer-GPU-2021.09.1.exe, RAM usage is around 4.5GB out of 11GB. That's with five 5mp cameras and four 2mp cameras using the BI optimizing guide.

With DeepStack-Installer-GPU-2022.01.1.exe installed and nothing else changed, RAM usage will eventually max out and Windows 10 becomes unresponsive and I will have to force restart the machine and cycle repeats.

I'm staying with DeepStack-Installer-GPU-2021.09.1.exe for now.
 
Top