PTZ5A4M-25X Metadata on NVR608H-32-XI

[Aquilles]

Hi everyone. I have a PTZ5A4M-25X linked to my NVR608H-32-XI and I disabled AcuPick in favor of metadata, and though my NVR supposedly support 8ch AI by NVR for metadata, it's not the case. Long story short I added two more color4k cameras that I need metadata by NVR but getting the "Reached max capability of AI analysis." only after 5ch.
I had initially done the the metadata by NVR for the PTZ but it seems this PTZ do support metadata as I see it's on the global config. How does the NVR pick this up? When I go to the Human Detection menu on the NVR and select AI by camera it says I need to add presets first though I have presets with IVS configured already.
When I look at the smartplan for the PTZ on the NVR is shows the global metadata is enabled.

 

[steve1225]

1. How many cams are on this system (NVR)? what type of cams?

2. Do you have enabled sub screen (separate video output for second/third/fourth HDMI) in video settings?
if yes - this eats around 40% AI power.. and VMD by recorder is going down to 5 channels...
If you read NVR spec, all AI features have different limits for Simultaneous mode and Heterogeneous mode...
Heterogeneous mode is when You have enabled separate screens for video outputs...

3. are the AcuPick are enabled on other cams?
if yes - each eats about 2% of AI power...

4. Do you have enabled any IVS / SMD / Face detection / Face recognition in 'AI by recorder' mode?
You must check on NVR ALL AI Parameters - go to IVS / SMD / Face detection / Dace recognition, iterate over EACH channel and check is somewhere on same channel is enabled 'AI by recorder' mode..

Each eats some of AI power...

 

[Aquilles]

Thank you for the reply.

1 x PTZ5A4M-25X
4 x Color4K-T
3 x IPC-T58IR-SE (planning on adding 7 more in the near future)

2. No subscreen enabled. I'm pretty sure it's Simultaneous mode, but for the life of me I can't find the location where it is again.

3. Yes the 3 x T58IR-SE have AcuPick and Face recognition enabled. I've already disabled face recognition which didn't helped on resources. Let me disable Acupick as well and check again.

4. Yes Face detection on 2x Onvif cams and VMD for the 4x 4K-T and 1x PTZ.
If only there is some indicator on how much AI resources is being pooled vs available and a breakdown of resources by feature or channel.

Any idea why the VMD of the PTZ is not feeding into the NVR?

 

[steve1225]

Do you want use video meta data done on camera or done on NVR for that ptz?

Show ai parameters -> video meta data for that ptz channel.

On main screen with icons click on right arrow and move to second page. There should be video icon where you have screen resolution settings. Left pane is for first output, right for the rest. It should be disabled.

Face detection and video meta data are using 1/8 ai power per channel. You use this for 7 channels so 7/8 of ai power is used. Last 1/8 is used somewhere else

Does you use face recognition? This also use ai resources

You can always default NVR to factory. Add that 7 plus one more camera where you want to run VMD on NVR, configure from scratch vmd / face detection, then add rest cameras.

 

[Aquilles]

Made no difference disabling Acupick. The reason I had Acupick enabled on T58IR-SE is that you get IVS and VMD by camera as opposed to only IVS when Acupick is disabled.
I went as far as disabling everything in AI parameters (heatmap, people counting, face detection, face recognition, ANPR) and still getting the max resource message. I only have 4 VMD by NVR configured on the 3x Color4K-T cams and the PTZ. The 4th Color4K-T is a no go.
Yep I agree perhaps a factory reset is needed given I've had the NVR configured and then did a major update recently. What a pain in the @$$ this is.

 

[steve1225]

Default to factory and redone all config on nvr will take max 1 hours.

Much faster that trying to resolve where the rest ai resources are going

 

[steve1225]

ps start adding cams from vmd ones. Configure vmd for them
Then face detection ones
Then AcuPick ones

 

[Aquilles]

Ok I did the default only. (All the parameters will be restored to default settings except network, user management and so on.)

Added all cams back in and without doing anything else, as you also suggested tried to setup VMD on the PTZ and 4x Color4K.
Same issue.

Going to go for factory default now, will let you know.

 

[Aquilles]

Nope not working. Factory reset, just added the PTZ and 4 Color4K-T cams back and tried to setup VMD and no matter what order you only able to setup 4 of them.

 

[steve1225]

What firmware you have on 608-Xî?

 

[steve1225]

Can you go to all color4k cam to image settings and reduce resolution from 4K to 4m (2688 × 1520)
change compression type to h.264H on both channels...

If you added any other ai done on NVR delete that.
Reboot nvr
And then try to add VMD to 5tf cam

 

[Aquilles]

Can you go to all color4k cam to image settings and reduce resolution from 4K to 4m (2688 × 1520)
change compression type to h.264H on both channels...

Thanks that actually worked. I played with the encoding and when all of them are on 4K and I drop one of them to 2592x1944 then I can add the 5th cam.
Is this expected behavior? The NVR is on firmware V4.000.0000000.6.R, Build Date: 2024-07-21

 

[steve1225]

OK - so the resolution was limiting...

If you look at specification this model have Decoding capability of 8 channels with 4k (8 MP) @ 30 fps with AI disabled and 4 channels with * 4k (8 MP) @ 30 fps with AI enabled..



You use AI, so for decoding / displaying over HDMI there is capability only for 4 channels with 4k (8 MP) @ 30 fps..

the missing decoding capability is reserver for AI use - like Video MetaData - You can use AI for cameras with max 4 channels with 4k (8 MP) @ 30 fps...

So this is why You couldn't add more cams to VMD by recorder..


Now after adding another camera with AI by Recorder, you can revert the full resolution on 4k camera and check what will happen..

go over building that every camera with VMD done by recorder saw You..

then check ai search->human does shows fresh records with You from all cams with VMD..

if this will work, then you've cheated the limiting system in the NVR and everything works..

if this will not work, then You must decrease frame rate on 4k cams to 15 or 20 fps..

ps. do you use live view over HDMI? in 3x3 grid?
if yes - then You also are limited by decoding capability (4 cams with 4k@30fps) - if NVR don't have enough capability it decodes / display secondary (low resolution) stream.

in that situation is worth decreasing frame rate on all 4k cams...

 

[Aquilles]

Now after adding another camera with AI by Recorder, you can revert the full resolution on 4k camera and check what will happen..

Ok so I did this already. Enabled AI on the camera, then went back and set the resolution back to 4K and it is working, showing attributes ect.

ps. do you use live view over HDMI? in 3x3 grid?

I always use web, hardly the monitor but I have live view set to a 4x4 grid but have now set this to 1. Thanks for pointing this out.

Thanks for the help. Really appreciate you narrowing it down. I have to admit RTFM.