I have been running deepstack in a container for over a year, but a few months ago it stopped working.   (I believe this may have been timed around when I switched from the old nVidia driver plugin to the newly created, official, nVidia driver plugin.)  The issue is that deepstack doesn't actually process the images it's being sent, and returns "nothing found" (this is the log info from Blue Iris that is sending the info).

 

 

I have Plex running on the same server, and it works with the GPU flawlessly so I don't believe there is any hardware problems.

I have tried the non-gpu version of the same container, with same settings (just edited the repository to drop the ":gpu") and it works poperly, albeit slowly.

I have installed deepstack-ui and it seems to timeout (10s timeout) when i upload an image.  The default image of the cyclist/car shows nothing detected.

when i visit the Web-UI for the deepstack container, it pops up showing deepstack is activated.

I have deleted the docker container and re-installed multiple times.
I have tried another repository (can't remember which one now, maybe a binhex?) as I wasn't using this current one before the problem showed up.  I switched as part of my quest to get it working again.

running "nvidia-smi" in the console does seem to show it running (python3 @ ~1500MB)

I have tried running with lower confidence, and turned on debug mode in the container (apparently sets confidence to 1 for everything).

 

 

root@Tower:~# nvidia-smi
Sat Apr 23 11:13:39 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.60.02    Driver Version: 510.60.02    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:02:00.0 Off |                  N/A |
| 42%   35C    P2    25W / 125W |   1538MiB /  6144MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      7797      C   python3                          1535MiB |
+-----------------------------------------------------------------------------+

 

unraid ver: 6.9.2

nVidia Driver: latest v510.60.02 (Have also tried production branch)

 

repository: deepquestai/deepstack:gpu

Nvidia Visible Devices: GPU-ba6c0608-289b-ac09-ecfc-9a9b6665abb6 (which i have confirmed via both the plugin page & "nvidia-smi -L")

Object Detection: True

 

I'm getting a bit stumped as to what to try next.
I'm starting to wonder if there is old data/config/etc laying around somewhere that I don't know about that is messing me up.
And just now I reverted back to CPU again and now that's not working...  banging head against wall....
EDIT: Rebooted server and the CPU version is back to working...
 

Edited April 23, 20224 yr by Xplode
code formatting

There are issues with the latest version of deepstack and I think its specific to the GPU version.  I saw posts where they acknowledged it and are looking into it.

 

I downgraded back to the previous version (2021.09.1) and it works fine.  Give that a shot until they figure it out.

 

Looking at my GPU while running it, the older version also uses much less memory.

 

I also noticed that blue iris is giving me issues when using custom models.  They are loaded in deepstack and all but BI just doesn't call the right URL's.  So I just use the default for now.

Ok thanks for the heads up on the version issue.   I'm not sure how to downgrade but I'll google that and hopefully sort it out.

I'm not using any custom models yet, so really just using it for person detection as I'm controlling some exterior lighting based on motion.

23 hours ago, Xplode said:

Ok thanks for the heads up on the version issue.   I'm not sure how to downgrade but I'll google that and hopefully sort it out.

I'm not using any custom models yet, so really just using it for person detection as I'm controlling some exterior lighting based on motion.

 

Just in case you can usually specfiy the vesion using the tag.  Sometimes its just a matter of adding ":2021.09.01-GPU".  Depends on the docker you are using.