Yes it is. I stopped the running deepstack, installed the new and restarted the new.
I am using the GPU version.
New DeepStack September Update Version 2021.09.1
- Thread starter MikeLud1
- Start date
I am using the GPU version.
I uninstalled the previous GPU version, deleted the folder, rebooted, then installed 2021.09.
It’s been running for ~24 hours now, and I’m noticing a increase in CPU usage on my system from ~21% to ~35%.
After another day I think I’ll reinstall the old pervious version to confirm that the apparent change is a consequence of the update.
Also, so far I’m not seeing a significant change in the confidence % in the confirmed alerts. A few users have reported an apparent improvement.
It’s been running for ~24 hours now, and I’m noticing a increase in CPU usage on my system from ~21% to ~35%.
After another day I think I’ll reinstall the old pervious version to confirm that the apparent change is a consequence of the update.
Also, so far I’m not seeing a significant change in the confidence % in the confirmed alerts. A few users have reported an apparent improvement.
Last edited:
I’m also not seeing any improvement in processing times -- compare the 'Mean' values in the following tables.
After Update:
Before Update:
Again this only after 24 hours, so I’ll continue the experiment for another day or two.
So far I must say I’m personally not noticing any tangible benefit after updating to 2021.09
(higher CPU usage and the same performance).
P.S. - For more background on the tables above, see this post
ipcamtalk.com
Some readers may notice that my processing times are higher than in my original post (~180ms vs ~140ms). This increase (almost 30%) is because I'm now using DeepStack mode 'High' instead of 'Medium'.
After Update:
Before Update:
Again this only after 24 hours, so I’ll continue the experiment for another day or two.
So far I must say I’m personally not noticing any tangible benefit after updating to 2021.09
(higher CPU usage and the same performance).
P.S. - For more background on the tables above, see this post
DeepStack Case Study: Performance from CPU to GPU version
I'm posting this thread for anyone considering making the transition from the DeepStack CPU version to the GPU version. Hopefully this will give you an impression of what you can expect in terms of enhanced performance. Please note some of my observations are system-specific. My system specs...
Some readers may notice that my processing times are higher than in my original post (~180ms vs ~140ms). This increase (almost 30%) is because I'm now using DeepStack mode 'High' instead of 'Medium'.
Last edited:
kklee
Pulling my weight
I updated the GPU version by just running the installer, didn't uninstall or change anything (although I took a full system image backup beforehand).
I ran a couple of missed detections through the live analyzer and it seemed to work better in that it detected what it should have.
It'll take some time to see if the overall detection has improved.
I ran a couple of missed detections through the live analyzer and it seemed to work better in that it detected what it should have.
It'll take some time to see if the overall detection has improved.
Ah, maybe this is benefit of updating?
But unfortunately very tedious to verify in a statistically significant way. To aid in this it would be helpful to be able flag false negatives in the cliplist as we notice them.
Ken recently added the ability to manually cancel and confirm alerts. Perhaps we can use a combination of the alert's confirm flag and memo value to flag false negatives.
Something like this...
/admin?camera=DW2&flagalert=3&memo=false%20negative ... unfortunately this command acts on only the latest alert.
Last edited:
I was a little concerned about CPU usage so I just renamed the old folder so I could revert back to the older version if I wanted to.
Below graph is approx. 30 minutes of CPU usage on the 'old' DeepStack install. 5 minutes in the middle with Blue Iris not running and then 30 minutes on the right with the latest Deep Stack. I should note -> I have several other things running on this computer - Plex, HomeSeer, and a couple of VMWare Workstation VM's. Currently running 16 cameras with 12 of them using Deep Stack during daylight hours (about 70 MP/s)
Below graph is approx. 30 minutes of CPU usage on the 'old' DeepStack install. 5 minutes in the middle with Blue Iris not running and then 30 minutes on the right with the latest Deep Stack. I should note -> I have several other things running on this computer - Plex, HomeSeer, and a couple of VMWare Workstation VM's. Currently running 16 cameras with 12 of them using Deep Stack during daylight hours (about 70 MP/s)
In post #22 I noted that I was seeing a CPU usage increase after upgrading to DeepStack GPU 2021.09.
Now I suspect this might have another cause. Yesterday I was having an issue. I restarted the Blue Iris service. My server's CPU usage then returned back to its pre-2021.09 level.
I want to amend the procedure I used to update DeepStack (post #22). I left out some details...
Now I suspect this might have another cause. Yesterday I was having an issue. I restarted the Blue Iris service. My server's CPU usage then returned back to its pre-2021.09 level.
I want to amend the procedure I used to update DeepStack (post #22). I left out some details...
- disabled DeepStack in Blue Iris... Settings > AI tab
- uninstalled the previous DeepStack GPU version in Add/Remove Programs
- deleted the DeepStack folder (after making a copy of my custom model folder)
- rebooted my server
- installed GPU version 2021.09
- added my custom model folder back to the new DeepStack folder
- started the new DeepStack Detection API in cmd.exe... deepstack --VISION-DETECTION True --PORT xxx
- enabled DeepStack in Blue Iris... Settings > AI tab
Last edited:
kc8tmv
Getting the hang of it
