AI is broken after switching from DeepStack to CodeProject

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,155
Reaction score
49,021
Location
USA
LOL, I was looking at the 2000-06-06 date :idk:
Oh crap LOL - yeah that is even worse LOL. Yeah if footage is questioned that is only a few seconds off, years off would be tossed so quick in court case. Wow.
 

actran

Getting comfortable
Joined
May 8, 2016
Messages
806
Reaction score
732
Yeah, camera timestamp is bad. @spazstic you should add NTP setup via cam webui to keep it correct.
 

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
His point is you have time coming from both the camera and from BI. Turn off the BI overlay. It may only be a fraction of CPU, but every little bit helps on an underperforming machine.

Plus since they don't match, should something happen that it were to go to court, the defendant attorney will have a field day discrediting the video, even if it is only a 3 second time difference. Or years in your case.

And who knows maybe the big @$$ BI time overlay is being perceived as motion and why you keep getting the camera triggered LOL. I have seen that happen on one of my cams so I excluded the time area from the Zone and it stopped the triggers.
The BI time/date is there because I'm viewing a past recording. It's not on the footage itself.

I know the time/date is incorrect. That's the first camera I purchased and it uses some proprietary plug in, in order to log into the camera. That plug in is now blocked in every browser I've tried. So - I can't make any changes to that camera. That camera isn't in a "critical" location so I'm not all that concerned about it. All the other cameras have the correct time/date.
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,155
Reaction score
49,021
Location
USA
Yet when this was first brought up you replied "Yes, why not? Let that overhead float on the camera side rather than the server."

Adding BI overlay to watch a past recording is adding undo CPU usage.

But lets move on - do your other cameras exhibit the same issue with CodeProject or just this one? Prior to when you couldn't get into this camera GUI, are you on default settings or did you tune it in for your field of view - settings of the camera can have a big impact on the success of AI. If this or other cameras are on default settings, that could be part of the issue.

The plug-in probably is probably for Internet Explorer - have you tried Internet Explorer - not Edge or Chrome with the IE tab but plain ole Internet Explorer?
 

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
@spazstic Look at the screenshot here. Make sure you have same settings. The list of custom models available to you is right there as a popup.

For each cam, set the custom model desired as shown. ipcam-general is good if you only need person/vehicle detection for both day and night. Look at other models for other objects. BTW, use higher confidence to eliminate false alarms. Again, CP.AI is using Deepstack models also so the quality of the predictions is driven by the model that you use.
This is great, thanks for the suggestion. I'll play around with the models and see if I can get an improvement.

However, I'm starting to think some of the issues may stem from the outdated hardware. CPU cycles to 100% often, and remains there during the object detection phase. I'm now looking into building a new PC with a 10+ gen Intel i5.
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,155
Reaction score
49,021
Location
USA
This is great, thanks for the suggestion. I'll play around with the models and see if I can get an improvement.

However, I'm starting to think some of the issues may stem from the outdated hardware. CPU cycles to 100% often, and remains there during the object detection phase. I'm now looking into building a new PC with a 10+ gen Intel i5.
An i5-8th gen is more than capable.

The other question is do any of your cameras have AI built in and would that be sufficient for your needs? If so your existing PC may be fine.
 

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
Yet when this was first brought up you replied "Yes, why not? Let that overhead float on the camera side rather than the server."

Adding BI overlay to watch a past recording is adding undo CPU usage.

But lets move on - do your other cameras exhibit the same issue with CodeProject or just this one? Prior to when you couldn't get into this camera GUI, are you on default settings or did you tune it in for your field of view - settings of the camera can have a big impact on the success of AI. If this or other cameras are on default settings, that could be part of the issue.

The plug-in probably is probably for Internet Explorer - have you tried Internet Explorer - not Edge or Chrome with the IE tab but plain ole Internet Explorer?
That overlay is only shown when watching a recording - I was referring to a permanent overlay on every video recorded...

I tuned the cameras for framerate/resolution/substreams. Not much else beyond that.

Yes, I tried IE as well. It's an old China camera but still works well, especially at night.
 

actran

Getting comfortable
Joined
May 8, 2016
Messages
806
Reaction score
732
@spazstic Yeah, PC with newer CPU will help alot if it is regularly pegged at 100%. Off loading AI to a Nvidia GPU will be good too. Check CP.AI website for list of supported GPU cards. If you do buy a new card, best to buy one with 4GB RAM or higher.
 

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
An i5-8th gen is more than capable.

The other question is do any of your cameras have AI built in and would that be sufficient for your needs? If so your existing PC may be fine.
The 10th gen would be helpful for native encoding on the Plex side of the server.

None of the cameras have AI worth a crap. They're inexpensive - mostly SV3C.
 

looney2ns

IPCT Contributor
Joined
Sep 25, 2016
Messages
15,646
Reaction score
22,918
Location
Evansville, In. USA
This is great, thanks for the suggestion. I'll play around with the models and see if I can get an improvement.

However, I'm starting to think some of the issues may stem from the outdated hardware. CPU cycles to 100% often, and remains there during the object detection phase. I'm now looking into building a new PC with a 10+ gen Intel i5.
No need to build.

 
As an eBay Associate IPCamTalk earns from qualifying purchases.

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
No need to build.

I looked into used HPs, but the proprietary MB mounting and PSU connections seemed more trouble than it's worth.

I have a RAID 6 card with 7x16tb drives mounted in a large case with enough cooling to keep the RAID card happy. Plus the 4tb drive for BI. Can't fit all that in one of those cases.
 
As an eBay Associate IPCamTalk earns from qualifying purchases.

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
@spazstic Yeah, PC with newer CPU will help alot if it is regularly pegged at 100%. Off loading AI to a Nvidia GPU will be good too. Check CP.AI website for list of supported GPU cards. If you do buy a new card, best to buy one with 4GB RAM or higher.
Rather than put money into a GPU with a power hog CPU, I'd rather build a decent system with a power efficient CPU and no GPU that can handle the task. Also, no GPU needed as I only remote in.
 

Coltect

Pulling my weight
Joined
Nov 3, 2017
Messages
55
Reaction score
129
Location
Australia
Rather than put money into a GPU with a power hog CPU, I'd rather build a decent system with a power efficient CPU and no GPU that can handle the task. Also, no GPU needed as I only remote in.
If you are looking at a newer system, i'd revert back to your old BI version and DeepStack on your current hardware, as you had that working well. Then get a new system and experiment with that until you have it working properly.
 

spazstic

n3wb
Joined
Oct 3, 2023
Messages
18
Reaction score
4
Location
Southern California
Built a new PC/server. Went from an old i7 970 to an i5-12400.

1Tb 990 Pro M.2 SSD, 32 Gb DDR4 RAM. Using old Sata SSD for New folder, with transfer to a 4TB WD purple after 2 days. I know this places a lot of wear on the SSD, but I don't really care. I have several of them I'm not using.

CPAI times went from 1-4 seconds to ~500ms. I got a P5000 for free so threw that in. Had to reinstall CPAI as CUDA drivers need to be installed prior to CPAI (took a minute to figure that one out).

With the P5000 I'm getting ~50ms average times. It's amazing.... Very fast response in BI, seamless scanning of substream video and alerts, and none of the issues with false positives that I had previously. I also reinstalled BlueIris, so that may have contributed, who knows.

Long story short - CPAI appears to put more load on the machine than DeepStack did, and will give false results if your CPU/GPU can't compute in a timely manner. I could have likely just upgraded the GPU to the P5000 (or a lesser GPU) and been fine. But the system serves other uses, and it was time for an upgrade anyway.... And my Plex users paid for all the parts :)

Thanks for all the help. :wave:
 
Top