5.5.8 - June 13, 2022 - Code Project’s SenseAI Version 1 - See V2 here https://ipcamtalk.com/threads/codeproject-ai-version-2-0.68030/
- Thread starter fenderman
- Start date
Try not running AI on all of your cameras if not needed. On my system I run 16 cams but only 7 use AI due to location and some inside. I also run a I7-6700k 16gb ram and it runs very well. But there are so many factors here that can affect the CPU usage.
With my aim of getting AI working, I've uninstalled DeepStack and Code Project AI then re-installed the latest Code Project AI. Revisiting the batch script install for CUDnn I have again run it as administrator. A window opens and 17 lines of "File system cannot find the path specified" and "press any key to continue", when I do the window closes. Does this represent a normal running of this batch file? Can I assume that CUDnn will have been properly installed and this is not the reason why AI isn't working here and I should look elsewhere for a reason it isn't?
MikeLud1
IPCT Contributor
Does your PC that you are running the scrip on have internet access? Because the script needs to download files. If the script ran correctly you should see the below folders
Last edited:
My many thanks MikeLud1 for all the hours of work you are devoting to this project.
Yes my BI PC has internet access, after all it installed Code Project AI and there were several files this needed to download and install.
I have 6 green boxes in presumable a working install of Code Project AI, from the browser in the BI's Global settings>AI.
I've been changing how clips are stored here, nearly 100% on New, zero on Stored, Alerts 2Gb. I have lost any clips from the green filing cabinet, the default, only if underneath I select New or Alerts these clips appear, but it's not permanent, nothing from Stored as I would expect. All my Alert clips have a no entry sign or the Ghostbusters logo in the top right had corner. I assume Alerts are being generated but not passed on to AI for filtering.
edit: sorry I needed to check. My contents of v8.5 look the same as yours. Perhaps I can conclude that CUDnn on my PC has been installed as expected.
Yes my BI PC has internet access, after all it installed Code Project AI and there were several files this needed to download and install.
I have 6 green boxes in presumable a working install of Code Project AI, from the browser in the BI's Global settings>AI.
I've been changing how clips are stored here, nearly 100% on New, zero on Stored, Alerts 2Gb. I have lost any clips from the green filing cabinet, the default, only if underneath I select New or Alerts these clips appear, but it's not permanent, nothing from Stored as I would expect. All my Alert clips have a no entry sign or the Ghostbusters logo in the top right had corner. I assume Alerts are being generated but not passed on to AI for filtering.
edit: sorry I needed to check. My contents of v8.5 look the same as yours. Perhaps I can conclude that CUDnn on my PC has been installed as expected.
Last edited:
MikeLud1
IPCT Contributor
To make sure Code Project AI is working run a test with the below link
Last edited:
Thanks, where is the file pexels-maria-orlova-4947381.jpg that I might use to test my Code Project AI install?
C:\Program Files\CodeProject\AI\TestData\Objects
Thanks, I found the file. In Custom Object Detection, Run Detection, General IP Cam it reports, "No predictions returned".
Being on 127.0.0.1 or local host makes no difference
I'm rather distracted, I started putting food out for a hedgehog a few days ago and I can seen on the CCTV we've now got two hogs simultaneously eating from the two bowls I put out. I'm half expecting a third to appear.
Being on 127.0.0.1 or local host makes no difference
I'm rather distracted, I started putting food out for a hedgehog a few days ago and I can seen on the CCTV we've now got two hogs simultaneously eating from the two bowls I put out. I'm half expecting a third to appear.
MikeLud1
IPCT Contributor
Run the attached Hardware Check and post what was displayed.
Well I been eyeing out this thread for 23 pages and finally decided to flip from DS to CodeProject.
I uninstalled DS, Cuda, cuDNN, deleted their respective directories. Then reinstalled Cuda 11.7 and then ran the script off the site for cuDNN and that went fine. Then installed CodeProject. Then just modified the registry like Mike showed earlier on in this thread, also modified the module settings json file to disable the stuff I didn’t want to see. Fired it up. Modified my model name on each camera settings and baam was back in business. I’m using a Quadro P1000 and it’s definitely doing better with this than DS was.
Just can’t seem to disable the Scene Classification module and the ObjectDetection (python). I’ve put false in basically everything except the Custom Detection one. It’s fine for now, cosmetic to me for now.
I uninstalled DS, Cuda, cuDNN, deleted their respective directories. Then reinstalled Cuda 11.7 and then ran the script off the site for cuDNN and that went fine. Then installed CodeProject. Then just modified the registry like Mike showed earlier on in this thread, also modified the module settings json file to disable the stuff I didn’t want to see. Fired it up. Modified my model name on each camera settings and baam was back in business. I’m using a Quadro P1000 and it’s definitely doing better with this than DS was.
Just can’t seem to disable the Scene Classification module and the ObjectDetection (python). I’ve put false in basically everything except the Custom Detection one. It’s fine for now, cosmetic to me for now.
MikeLud1
IPCT Contributor
In the vision folder the modulesettings.json has more then one Activate to set to False, scroll down and look for all of them.
My results of the hardware check on the PC dedicated to Blue Iris is here. Seems to be results showing my T600 and my EliteDesk 800 G3 i7-7700's onboard Intel 630 graphics.
The T600 seemed to be the highest performing GPU that this PC could take considering its consumption wattage.
Last edited:
I wrote, "I have lost any clips from the green filing cabinet, the default, only if underneath I select New or Alerts these clips appear, but it's not permanent...".
After an overnight crash of BI on 5.6.0.7 I rolled back to 5.5.9.6, it's returned. A click on the filing cabinet turns it green and I have a list of all video clips back instead of a black empty box down the side. The list of Alerts is still available if I click on the lightening flash icon as usual, although all the Alerts are still showing at the top right the white circle with bar, AI is still not working here. I thought I had caused the All clips to be unavailable because I've been changing BI settings etc hoping to get AI working.
So my usual experience of viewing lists of clips is back to normal on 5.5.9.6 (1st August).
Because Code Project AI isn't working here I've been willing to install the latest BI releases, that though comes with consequences I didn't expect. On 5.5.9.6 I'll also be interested to see how stable BI is on this, it was stable in the past.
My question to Ken...
I’m still running Deepstack but would like to move to CodeProject.AI. Right now, there’s a bit too much manual editing the registry etc. to get things the way I’d like. Do you expect more work to be done on making CodeProject.AI more configurable from within Blue Iris?
Ken's answer...
Yes, absolutely we are working with the CodeProject.AI developers on making everything even more seamless.
I’m still running Deepstack but would like to move to CodeProject.AI. Right now, there’s a bit too much manual editing the registry etc. to get things the way I’d like. Do you expect more work to be done on making CodeProject.AI more configurable from within Blue Iris?
Ken's answer...
Yes, absolutely we are working with the CodeProject.AI developers on making everything even more seamless.
Village Guy
Pulling my weight
@MikeLud1
Further to your explanation and recommendation to use low resolution streams for AI analysis. I'm having difficulty understanding how any software can analyse a snapshot of such low resolution for face identification. By way of example, I have attached a capture from my doorbell camera 640 x 480 of visitors. Needless to say, the face is unrecognised even though a good quality capture in higher resolution was in the library. I'm not sure that I could recognise the person from the image captured. The only way I could see face recognition working is for BI to crop just the face and send it for analysis at the original resolution which could be actually end up with an image less than 640 x 480.
Am I missing something fundamental? I would also be interested in your thought on the image processing times which seem excessive to me.
Further to your explanation and recommendation to use low resolution streams for AI analysis. I'm having difficulty understanding how any software can analyse a snapshot of such low resolution for face identification. By way of example, I have attached a capture from my doorbell camera 640 x 480 of visitors. Needless to say, the face is unrecognised even though a good quality capture in higher resolution was in the library. I'm not sure that I could recognise the person from the image captured. The only way I could see face recognition working is for BI to crop just the face and send it for analysis at the original resolution which could be actually end up with an image less than 640 x 480.
Am I missing something fundamental? I would also be interested in your thought on the image processing times which seem excessive to me.
