CrazyAsYou
Getting comfortable
Anyone else wish the forum had an option to disable pagnation and just have infinate scroll/load, would be great for posts with 52 pages!
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
biggen
Known around here
- Joined
- May 6, 2018
- Messages
- 2,567
- Reaction score
- 2,842
Yeah that would make it easier to read for sure. I spent about an hour skimming this thread thread before I took the plunge. We do need a FAQ or small how-to guide but the project is changing so rapidly it could be made irrelevant quickly upon changes.
Vettester
Getting comfortable
- Joined
- Feb 5, 2017
- Messages
- 743
- Reaction score
- 698
Last edited:
I installed the latest BI version and downloaded CodeProject on a test computer. Most every thing is working very well.
USPS is triggering when a vehicle is detected and LPR is not working.
USPS is triggering when a vehicle is detected and LPR is not working.
Last edited:
MikeLud1
IPCT Contributor
The LPR is not released yet, I am still working on it.
fenderman
Staff member
- Joined
- Mar 9, 2014
- Messages
- 36,902
- Reaction score
- 21,274
This is going to be great.
Mine has been massively more stable than it was with DS over the past few weeks, with DS it would require a reboot (sometimes a forced power off) at least once every 3-4 days (if not more often) with it just going into timeout constantly or crashing the OS. I thought there was a hardware fault on my pc but since CPAI not a single issue. Looking forward to the LPR being added 
Corvus85
Getting the hang of it
Same. DS was a POS. Every couple weeks, it just refused to detect anything and become non-responsive, even though the test web page worked. The only thing that fixed it was reinstalling DS, which was a royal PITA.
CPAI has worked flawlessly ever since I installed it, and with much better response times/lower CPU usage. I'm keen to see how it improves even further.
I’m using Object detection (YOLO), put delivery.pt in the custom models folder but it is not being processed. Any ideas?
MikeLud1
IPCT Contributor
Post screenshots of the Main AI settings and the Camera AI settings
MikeLud1
IPCT Contributor
Do you have both Object Detection (.NET) and Object Detection (YOLO) enabled? If so you can only have one enable not both.
I can activate. Net and Yolo all working... Crazy
Okay now I can only activate one.. Net with gpu. But Yolo only cpu. I have disable half. My GT 1030 not working with half.. But hardware type gpu
Attachments
Last edited:
cybernetics1d
Getting the hang of it
For those that use CPU-based detection, do you use .NET or YOLO option? I believe if you're using custom models, then YOLO is the only option, but I'm curious how much performance differences between the two option. Has anyone done extensive benchmark between the two? Previously, I was using Deepstack in container with P400 GPU and was seeing ~100ms performance and now with latest version of CodeProject.AI, i'm seeing 120-200ms performance with ipcam-general custom model, which is somewhat similar to what I've seen before running Deepstack CPU. Surprisingly, I'm only seeing very little CPU spike when comparing to Deepstack previously.
I am about ready to take the plunge and upgrade from Deepstack to CP.AI. I'd appreciate any advice or comments on the proper approach.
Currently running an I5 Windows 10 system with a Quadro P400/2GB DDR5. BI is 5.5.7.11. (I installed an older version of CP.AI but it didn't work right and I backed down until I could focus on the project. BI is older because I had Deepstack issues with newer versions of BI.) CUDA is v10.1.243.
I have downloaded the CP.AI script "install-CUDnn.bat" and CUDA Toolkit 11.8 and CP.AI 1.6.7.0. I also made a complete C: drive image backup so I can revert if it all goes badly.
My plan is to:
1. Upgrade the P400 drivers to Release 515 (following the instructions in the script to upgrade to 11.7 drivers, 11.8 is the latest version that nVidia shows).
2. Upgrade the CUDA Toolkit to 11.8. Is there any reason to upgrade specifically to 11.7 even though 11.8 is available?
3. Run the "install-CUDnn.bat" script
4. Install CP.AI v1.6.7.
5. Test to see if anything yet has broken the current Deepstack just so I know have a still working system to which I can return if needed. Probably do another C: image backup.
6. Update BI. Choices are stable 5.6.1.3 or beta 5.6.2.9. Is the latter needed for CP.AI?
7. Point BI to CP.AI instead of Deepstack. Do more research into custom models (Thanks @MikeLud1 !) and fine-tune.
Please tell me if there's something stupid in the above.
Currently running an I5 Windows 10 system with a Quadro P400/2GB DDR5. BI is 5.5.7.11. (I installed an older version of CP.AI but it didn't work right and I backed down until I could focus on the project. BI is older because I had Deepstack issues with newer versions of BI.) CUDA is v10.1.243.
I have downloaded the CP.AI script "install-CUDnn.bat" and CUDA Toolkit 11.8 and CP.AI 1.6.7.0. I also made a complete C: drive image backup so I can revert if it all goes badly.
My plan is to:
1. Upgrade the P400 drivers to Release 515 (following the instructions in the script to upgrade to 11.7 drivers, 11.8 is the latest version that nVidia shows).
2. Upgrade the CUDA Toolkit to 11.8. Is there any reason to upgrade specifically to 11.7 even though 11.8 is available?
3. Run the "install-CUDnn.bat" script
4. Install CP.AI v1.6.7.
5. Test to see if anything yet has broken the current Deepstack just so I know have a still working system to which I can return if needed. Probably do another C: image backup.
6. Update BI. Choices are stable 5.6.1.3 or beta 5.6.2.9. Is the latter needed for CP.AI?
7. Point BI to CP.AI instead of Deepstack. Do more research into custom models (Thanks @MikeLud1 !) and fine-tune.
Please tell me if there's something stupid in the above.
MikeLud1
IPCT Contributor
Stick with CUDA 11.7
Thank you, @MikeLud1
I did as you suggested, and to my utter amazement, everything seems to be working. I have a lot of work to understand and fine-tune the trigger sensitivity and custom model selection, but basic functionality is there.
Enable. Net and Yolo same
Attachments
MikeLud1
IPCT Contributor
Check your log, see my log below, also when I have both enabled the custom model does not work right see below, the ipcam-general model does not have potted plant.
2022-10-24 03:33:48: Server: Started Object Detection (.NET) backend in Server
2022-10-24 03:33:51: Server: ObjectDetectionNet.exe: Application started. Press Ctrl+C to shut down. in Server
2022-10-24 03:33:51: Server: ObjectDetectionNet.exe: Hosting environment: Production in Server
2022-10-24 03:33:51: Server: ObjectDetectionNet.exe: Content root path: C:\Program Files\CodeProject\AI\AnalysisLayer\ObjectDetectionNet in Server
2022-10-24 03:33:52: Object Detection (Net): Object Detection (Net) module started. in Object Detection (Net)
2022-10-24 03:33:52: Server: ObjectDetectionNet.exe: Please ensure you don't enable this module along side any other Object Detection module using the 'vision/detection' route and 'detection_queue' queue (eg. ObjectDetectionYolo). There will be conflicts in Server
CCTVCam
Known around here
- Joined
- Sep 25, 2017
- Messages
- 2,674
- Reaction score
- 3,504
What's the position on SenseAi vs DS now?
Also, what's the position with GPU vs CPU?
I know DS was recently reported as having made GPU@s effectively obsolete for detection and CPU alone was sufficient yet from the bit I can skim read on recent posts, is SenseAI still reliant on GPU for good performance / fast detection?
Although I love a development thread like this and it's interesting, unless you follow the thread from the very start, it canbecome very confusing with info split over dozens of pages. There's sometimes a lot to be said for separate thread which is trimmed down to just give the basics - what it is, how to install, how to use, latest changes (maybe incorporated into updated usage instructions or feature list).
Also, what's the position with GPU vs CPU?
I know DS was recently reported as having made GPU@s effectively obsolete for detection and CPU alone was sufficient yet from the bit I can skim read on recent posts, is SenseAI still reliant on GPU for good performance / fast detection?
Although I love a development thread like this and it's interesting, unless you follow the thread from the very start, it canbecome very confusing with info split over dozens of pages. There's sometimes a lot to be said for separate thread which is trimmed down to just give the basics - what it is, how to install, how to use, latest changes (maybe incorporated into updated usage instructions or feature list).