IP Cam Talk Custom Community DeepStack Model
- Thread starter MikeLud1
- Start date
see
Deploying Your Model With DeepStack
docs.deepstack.cc
Sorry, I should have clarified further.see
and for the BlueIris config see the first pages of this thread, but I found a problem with that (a couple of posts back)Deploying Your Model With DeepStack
I'm running DS in Docker in a VM on a separate computer with VorlonCD's bi-aidetection. Not using BI DS.
Thanks!
Hello Mike, I'm fresh into the surveillance system game with 2 5MP IP cams running Blue Iris and DeepStack. Where do I get these custom models? Can anyone use these models? Can I contribute to this project? Please let me know and thanks in advance!
In first post
IP Cam Talk Custom Community DeepStack Model
I figured out how to download the images that DeepStack used to train the object model. With these images I can start to make the community custom DeepStack model, below are the steps to create the custom model. The first step would be to create a new DeepStack custom model using the same image...
Yes
At this time I am taking a break on this project, I will resume shortly. I have been working on the below ALPR project.
Blue Iris and DeepStack ALPR
I took a break from my DeepStack Custom Model project and started work on a DeepStackALPR solution. I made a lot of progress on the DeepStackALPR. The solution is API that crops and rotates (if needed) the license plate. Then the API reads all of the characters in the plate and logs the license...
Thanks for the quick reply! So I've been having issues with high CPU usage when DeepStack is analyzing. I hear of people having dozens of cameras on equipment that's a lot older than mine and they have like 500ms process times. I only have 2 - 5MP Amcrest cameras running 15 fps with substream and have done all of the optimizations recommended in the forum. As soon as the camera is triggered my CPU goes straight up to 100% utilization. I have a Core i7-11700k processor so that should not be the issue but I can't think of what else would be causing so much usage. Do you have any tips?In first post
IP Cam Talk Custom Community DeepStack Model
I figured out how to download the images that DeepStack used to train the object model. With these images I can start to make the community custom DeepStack model, below are the steps to create the custom model. The first step would be to create a new DeepStack custom model using the same image...ipcamtalk.com
Yes
At this time I am taking a break on this project, I will resume shortly. I have been working on the below ALPR project.
Blue Iris and DeepStack ALPR
I took a break from my DeepStack Custom Model project and started work on a DeepStackALPR solution. I made a lot of progress on the DeepStackALPR. The solution is API that crops and rotates (if needed) the license plate. Then the API reads all of the characters in the plate and logs the license...
Post a screenshot of one of the camera's AI settings and the main AI settings. Are you using the custom models in the first post?
Which version of DeepStack did you install? Did you use the GPU version? If you do not have an Nvidia GPU, you may see CPU spikes.
Last edited:
I am using the GPU version of DeepStack. If you have a Nvidia GPU then you should use the GPU version, that will increase detection times and reduce CPU utilization.
I'll send screenshots when I get to my computer. I am using the CPU version. I did just buy an RTX 2060 graphics card in hopes to get that usage down. I've heard great things about it so fingers crossed. And I am using the "combined" custom model.
With the RTX 2060 you will see 100ms +-25ms and no impact to your CPU.
That sounds perfect to me. I'll currently get 6000ms+ times way too frequently and I can't figure out why. Like I said, I'll send screenshots when I'm home and am just hoping that I have some settings wrong. Thanks for the feedback.
Here are the screenshots of my camera's settings. I've tried using only 5 real-time images at 1-second intervals and I got the same results.
Attachments
You should be using the below settings for the Custom models. Your current setting is running three models the default deepstack model, combined, and general
sebastiantombs
Known around here
I'd cut the number of images used down to maybe 10. 999 is really high and will keep the CPU churning for the full length of a clip, at 999 with a 500ms timing that means 45.5 seconds worth for a trigger.
Even when I just had 5 images, 1 per second my CPU went up to 100% immediately. Nothing seems to keep my CPU usage to a reasonable level.
I JUST ordered it yesterday so I should see it before the weekend. I'm still getting all of my necessary alerts, it's just taking a lot more of out my CPU than I expected. Just recently I got an 8654ms process time...I've attached a screenshot for referance.
