Anyone seen this?

I wrote about this cam 2 weeks ago, there is demo video from Polish Dahua channel:

It generates 3 separate channel streams...
IVS is working on each channel, but it looks like the configuration is shared for all 3 channels (You draw IVS rules on wide view and camera rescales the ones which fits into the second/third zoom/telezoom view)..

Firmware is based on riemann - so it's the same SOC (and video processing) as in 5xxx-S3..
Frame rate is limited to 25fps..
All AI functions (except IVS) are not present (no SMD, no AI Encoding, no AI SSA, no AcuPick, no VideoMetaData)..
It looks like they use the SOC at maximum power capacity (3 separate channels, 3 IVS)..


Apologies to you and @Mark_M for re-posting as I didnt see your post. Always appreciate your insights, thank you guys
 
  • Like
Reactions: JDreaming
We are working on that PTZ with 180 camera video soon. Should be ready at early Oct.
 
  • Like
Reactions: JDreaming
It generates 3 separate channel streams...
IVS is working on each channel, but it looks like the configuration is shared for all 3 channels (You draw IVS rules on wide view and camera rescales the ones which fits into the second/third zoom/telezoom view)..


Confirmation that IVS configuration is shared between channels:

Screenshot 2024-09-19 at 14.00.16.png


There is mini vertical PTZ for angle adjustment:

Screenshot 2024-09-19 at 14.01.15.png
 
If we talk about solutions for stadiums..

Panomera cams starts at 32Mpx with 19K Euro price tag and some fancy multi sensor shit..

Axis have normal bullet, 41Mpx with one 4/3 sensor and 21-7 degree horizontal FoV lenses and 5K USD price tag..

I have two more question:
How to display 190MP video feed? 8K monitors/TVs are 32MPx...
As I understand they have specialized VMS software with multiple EPTZ functionality..

How to search for a specific person among the crowd of thousands recorded by this cam??
The concept of multiple imaging sensors is to maintain a high pix/inch at various distances.
The Panorama cameras are setup to factor depth/distance of each point within the picture. Ideally to a target of 250pix for a face per seat.

EPTZ type viewing and powerful serves/computers. It's a system to watch 'after the event' or digitally zoom in by an operator as they spot something.
Given that this sees an entire crowd, I think we'll see more of these for weapons detection AI.
 
  • Like
Reactions: bigredfish