CodeProject.AI Version 2.0
- Thread starter MikeLud1
- Start date
biggen
Known around here
- Joined
- May 6, 2018
- Messages
- 2,609
- Reaction score
- 2,926
Probably some expensive NVR only solution.
MikeLud1
IPCT Contributor
The current version is 2.3.4, version 2.4.x should be released shortly, I am currently prerelease testing.
MikeLud1
IPCT Contributor
Below are details of an upcoming release
MikeLud1
IPCT Contributor
I will try to put something together this weekend
Are you saying the BI will use 2.3.4? I'm on 2.0.6 right now as one of the older BI version had not supported the latest cp-ai at the time; nothing was being detected. I'm not using the license plate option.
MikeLud1
IPCT Contributor
BI has been working fine for me on v2.3.4. Version 2.4.x should be released soon. So wait to upgrade until v2.4.x comes out. When upgrading you will need to do a full uninstall before upgrading.
Same for me did you ever figure out what it was?
Sorry to jump in here & may have been mentioned but this seems a very knowledgable thread.
Wanting to speed up the AI process.
I use motion detection then AI (now CPAI, just changed from deepstack) to make sure a vehicle ,
then I use python code to do the LPR (plate recogniser SDK) then send telegram to advise X is at the gate & if known it opens the gate. What I would like is for the gate to open almost immediately however currently can take from 4 to 10 seconds, not that much of a big deal really but seems unnecessarily slow
My code is fast, less than 1 seconds, it just the AI I think is slower than I would like. 2nd hand i7 4760 16gb, I thought pretty fast albeit rather old. Handles motion & 7 cameras great. Seems from motion detection to the alert is 4-10 seconds
Have ordered a RX6400 so hoping this will give quite a boost, always go Nvidia normally but limited choice, slim case & low power, OEM HP PC.
What I am trying to work out is does the AI analyse every image even if the first one it finds is a car, I can see all the ms data for following images so guess it does but strangely does not seem to take that long, I would think much more than 10 seconds if that was the case.
I understand if you want a cancel option it would have to check all images before alerting but if nothing in "cancel" can it or does it alert immediately.
I want it to potentially detect a car early on to be fast but if poor light etc. so I need 20'ish+ real time images to give it a chance to be sure of a car as it gets closer. Slow detection is fine in this situation but 90% of time catches "car" very early on & good to make use of that.
sorry bit rambling in the end.
Wanting to speed up the AI process.
I use motion detection then AI (now CPAI, just changed from deepstack) to make sure a vehicle ,
then I use python code to do the LPR (plate recogniser SDK) then send telegram to advise X is at the gate & if known it opens the gate. What I would like is for the gate to open almost immediately however currently can take from 4 to 10 seconds, not that much of a big deal really but seems unnecessarily slow
My code is fast, less than 1 seconds, it just the AI I think is slower than I would like. 2nd hand i7 4760 16gb, I thought pretty fast albeit rather old. Handles motion & 7 cameras great. Seems from motion detection to the alert is 4-10 seconds
Have ordered a RX6400 so hoping this will give quite a boost, always go Nvidia normally but limited choice, slim case & low power, OEM HP PC.
What I am trying to work out is does the AI analyse every image even if the first one it finds is a car, I can see all the ms data for following images so guess it does but strangely does not seem to take that long, I would think much more than 10 seconds if that was the case.
I understand if you want a cancel option it would have to check all images before alerting but if nothing in "cancel" can it or does it alert immediately.
I want it to potentially detect a car early on to be fast but if poor light etc. so I need 20'ish+ real time images to give it a chance to be sure of a car as it gets closer. Slow detection is fine in this situation but 90% of time catches "car" very early on & good to make use of that.
sorry bit rambling in the end.
veesion doesn't use an expensive nvr, actually u need cheaper stuff (i have 3 different systems on my supermarkets and only the cheapest was compatible with veesion)
they give u a "server" that u have to add on ur camera network and this will do the checks, with a few problems:
i asked for a demo -> no demo
atleast a year of sub + 1k for buy the "server" (probably a raspberry or something worth 100-200€)
i know another supermaket owner near me who bought that sw, i asked if he would rebuy it and he said yes, but he isn't the one who use it, so i don't think is an accurate answer
from what i know it gave you a lot of false positive, i didn't personally saw the system/results, but if you get an allert for every customer who put the phone in the pocket is pointless
last thing make a video where the camera is at 1mt from the thief is always easy
they give u a "server" that u have to add on ur camera network and this will do the checks, with a few problems:
i asked for a demo -> no demo
atleast a year of sub + 1k for buy the "server" (probably a raspberry or something worth 100-200€)
i know another supermaket owner near me who bought that sw, i asked if he would rebuy it and he said yes, but he isn't the one who use it, so i don't think is an accurate answer
from what i know it gave you a lot of false positive, i didn't personally saw the system/results, but if you get an allert for every customer who put the phone in the pocket is pointless
last thing make a video where the camera is at 1mt from the thief is always easy
Last edited:
wpiman
Pulling my weight
Is it possible for you to move the camera further up the road before you get to the gate? My LPR in on my street and I have similar delay times...
prsmith777
Getting comfortable
Release Notes - CodeProject.AI Server v2.6.2
A Guide to using and developing with CodeProject.AI Server
www.codeproject.com
there are release notes for 2.4.1 but I cant find a download link for that version.
MikeLud1
IPCT Contributor
The new version should be released any day now, there where some bugs. The version I am testing is 2.4.3Release Notes - CodeProject.AI Server v2.6.2
A Guide to using and developing with CodeProject.AI Server
there are release notes for 2.4.1 but I cant find a download link for that version.
I also have been testing the 2.4.3 CP beta using the new "Mesh" feature. I use CP in the CPU mode only in a windows W11 environment so it has been very stable for me. The mesh timings are very good. My office computer is only on during the day and early evening and then gets shut off every night. When I start it up every morning my BI machine will then see the mesh slave is available and starts using it when needed. Its all automatic and has been running great! This is an awesome feature in my case.
Vettester
Getting comfortable
- Joined
- Feb 5, 2017
- Messages
- 751
- Reaction score
- 709
The new Mesh feature uses multiple computers running CPAI to lower the processing time for the Obj Det?? So for instance you could be running 5 BI servers with 20 cameras each, all running CPAI all sharing each others CPAI service? Or would that be a wrong way to look at it?