MikeLud1
IPCT Contributor
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
Vettester
Getting comfortable
- Feb 5, 2017
- 959
- 936
Thanks Mike
Will give it a go and report back
I added the change from DeepStack for completeness, as can be seen there is definitely the potential for SenseAI to go faster.
Will give it a go and report back
I added the change from DeepStack for completeness, as can be seen there is definitely the potential for SenseAI to go faster.
| Detection Times | Number of samples | ||||||||
| DeepStack (ms) | 1.6.8.0 (ms) | 2.0.6 (ms) | Note | DeepStack (ms) | 1.6.8.0 | 2.0.6 | |||
| Garage | 343.4 | 265.9 | 507.4 | 4k | Garage | 1,193 | 906 | 166 | |
| House | 313.9 | 124.3 | 503.4 | Substream | House | 840 | 642 | 154 | |
| Porch | 311.3 | 147.5 | 531.2 | Substream | Porch | 478 | 420 | 107 | |
| Shed | 232.8 | 117.9 | 296.2 | Substream | Shed | 769 | 157 | 160 | |
| Overall | 305.2 | 188.7 | 446.5 | Total | 3,280 | 656 | 1,838 |
Thanks Mike
Using that file and a few tests the times were in the 170-180ms range so a huge improvement over the >500ms most cameras took on 2.0.6.
Not quite as good as 1.6.8.0 so far but that was only a handful of tests in light that is failing but not quite reduced enough for the cameras to flip to B&W so I'll see how this is looking after a couple of weeks normal use.
Thanks for the help
Using that file and a few tests the times were in the 170-180ms range so a huge improvement over the >500ms most cameras took on 2.0.6.
Not quite as good as 1.6.8.0 so far but that was only a handful of tests in light that is failing but not quite reduced enough for the cameras to flip to B&W so I'll see how this is looking after a couple of weeks normal use.
Thanks for the help
CrazyAsYou
Getting comfortable
I see this dropping the resolution down from 640 to 416. I assume this is only any good if you're using medium models which should be 416 but would make no difference to us using larger models @ 640?
I installed codeproject 2.6 on an server with GPU, currently running a blueiris evaluation. On a good note, codeproject sees that I have a GPU. Quick recap, I coped the configuration that is working from the other server and it coped successfully all the values, cameras and etc. Unfortunately while I'm getting the alert on the other server no alerts on this one. In codeproject I'm getting this message. Timeout connecting to the server, alghouth the codeproject shows that is up and running using GPU and CPU. At one point I disabled the GPU without any effect to the alerts.
Thank you, this is a great group by the way.
Sent from my SM-S906U using Tapatalk
Thank you, this is a great group by the way.
Sent from my SM-S906U using Tapatalk
I have detections set using ipcam-combined.pt and delivery.pt. My detections relying on the delivery.pt model never worked - everything was detected as FedEx. Since upgrading to the 2.06 beta, I get the correct indications for UPS, USPS, Amazon, etc. I did not make any other changes.
Can someone point me in the direction of how to use it please? I'm not sure what settings to use on the following screen. Also, do we have access to the plate as a variable to be able to pass to an API?
Thanks
Vettester
Getting comfortable
- Feb 5, 2017
- 959
- 936
I didn't change anything on the way mine is setup, but I don't think it is configured correctly. I got 1 license plate when the app first started and I enabled the LAPR option, but now all of my dayplate captures are going into the canceled alerts folder.
@MikeLud1 - Mike would you please give us some guidance on how to optimize this?
Edit: The CPAI Server seems to be capturing the plates.
@MikeLud1 - Mike would you please give us some guidance on how to optimize this?
Edit: The CPAI Server seems to be capturing the plates.
Last edited:
MikeLud1
IPCT Contributor
Below are my setting, I am still confirming on how Ken setup the integration.
In the camera AI setting I am using "t" so the AI runs all 4 real-time images
Vettester
Getting comfortable
- Feb 5, 2017
- 959
- 936
Last edited:
@MikeLud1 - it almost appears like we should have a new thread started that consolidates/streamlines the latest revision?
I know that no matter what point a thread like that starts, there always comes that time with dated information, but a lot has changed since this thread first started and 79 pages later and I think a lot of people will be making the switch now that is appears a lot of integration is in place?
I know that no matter what point a thread like that starts, there always comes that time with dated information, but a lot has changed since this thread first started and 79 pages later and I think a lot of people will be making the switch now that is appears a lot of integration is in place?
When I click the custom models button I keep getting a message to restart the AI to refresh the list. I restarted BI and also the AI service but still keep getting the message. Any ideas?
MikeLud1
IPCT Contributor
mine says "Installed Modules:" (nothing shows) "restart AI to refresh this list" (tried stop/start many times) CPAI .NET shows running
MikeLud1
IPCT Contributor
MikeLud1
IPCT Contributor
I agree, @fenderman do you agree we should start a new thread named CodeProject.AI Version 2.0