Subscriber Discussion

Analytics On Demographics?

GG
Gautham Gopalakrishna
Sep 11, 2013

Anyone has used 3VR's demographics analytics? How effective it is in terms of accuracy? I'd like to understand from end user perspective. Anyone else offering a similar solution?

JH
John Honovich
Sep 11, 2013
IPVM

We have not used it but we got a briefing and demo - Here is our overview of 3VR's demographic analytics.

Others with offerings: Cognitec Demographics and Panasonic Business Intelligence Kit

Avatar
Darren Ruger
Sep 11, 2013

Hi there Gautham,

Yes, we use the 3VR suite of products extensively down here in Australia to various degrees of success. The first issue to overcome is how the analytics are actually set up (this is not a simple “plug and play”).

My team and I actually flew up to San Fran to do the weeks training and then stayed on to work with the engineering team up there to understand the finer idiosyncrasies of the software and our “real world” discoveries with regards to the software not performing were ALL based around the camera selection and set up (and particularly FOV).

These analytics are very sensitive, and our findings are that they can be very accurate when all of the elements are deployed correctly.

JH
John Honovich
Sep 11, 2013
IPVM

Thanks, Darren!

Was it primarily than angles to subjects and width of FoV? Or were there other core drives for optimizing?

Avatar
Darren Ruger
Sep 11, 2013

Angle of declination is critical but if using Megapixel cameras the width of the shot came into play quickly with the “optimised” area of detection being 640x480 so there is sometimes a need to set up multiple detection zones in a single shot.

KG
Kison Ghan
Sep 12, 2013

I have not seen 3VR in actions, but in normal casre, Demographic Analytics (DA) ranges from 70% to 80% (with +/-10%), and that accuarcy fluctuate between between Age, Gender and Race, any claims above 80% accuaracy are likely to come with narrow environment directive which again will not increase the accuarcy to much due the broadness of race, skin colour, bone-structure and other combinations, each race with their specific elements has to be specificly trained with at least 10,000 pictorial sample before it even began to be use-able.

Again, what are purpose of the customers? General data-mining or mission critical.

DA is a processor intensive application, it does not comes embedded with cameras, Low-End version normally for "advertising attention rating" comes bundled with low-end PCs, High-End DA (security and retail analytics) will definitely requires sever class processing environment.

Message me if you requires specific application for DA.

New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions