BI with AI on a seperate machine, cannot use face rec.

Joined
Apr 5, 2022
Messages
3
Reaction score
0
Location
Minnesota
Hello,
I have a BI cam setup with about 36 cameras on a server 2019.

It is hosted on a 4 node cluster (each cluster is dual xeon and 96G ram and 2 10G fiber ethernet), I give BI 64GB ram and 24Cores of xeon at 90%.

The BI system works very well and pretty stable, every once and a while get an issue, usually me messing with it or things. Handles the live cutovers well.

I currently host my AI on a non cluster machine, I dont have 4 the same gpu to put in the cluster, so I run on an external machine.

This works a treat, usually YOLO requests take 20-30ms every once and a while i get one that takes a bit over 100ms, which work good, the BI system will generate alerts and put the clips in the db, everything works a treat.

Except...

When i turn on face recognition, I dont get anything, NO unknown faces in the folder, nothing at all.

I had enabled it in the past, but was getting a bit much processing everything on cpu, costing a lot of cpu time, this is the reason i did the external system running the ai.

I also just on an off chance mapped the same V: drive exactly the same on the AI server as it is on the BI server (Crap shoot...).

But for some reason I do not get any faces in the face directory, and I know there are plenty of faces to look at on the cameras.

I have the selection checked to save the unknown faces to a directory on the V: drive (same drive/map as the main clips, just different folder)

Anyone have this situation, and/or even have a system to test/see if there is something in maybe my setup???

Thanks
LKY

P.S.
I do have clients running BI that may ask about this feature, so im checking it out.

I can get remote access if needed to this system, i consider it a test system, as most are recorded elsewhere as well on panasonic dedicated hardware, this is more of a convience to have BI running and nice to use the app.

Most my clientss are still on Milestone as of yet, but slowly trying to save them a few bucks and get them some features.
 
Top