Dr Hartmut Neven Founder (left) Alex Cory CEO (right)
A recent campaign using a movie poster logo.
A recent mobile marketing campaign in Germany using the TV and mobile.
8 comments:
Anonymous
said...
Scott What differences are there between Nevenvision and Mobot. Did you get a chance to compare both at the show? By the way how did Ironman go for you? Larry
The movie poster example simply reinforces a point I made in a post elsewhere on this blog: how well can this technology possibly scale up?
How many movie posters are there? How many ads in ad campaigns? How many company logos? Tens of thousands? Hundreds of thousands?
How do you get a crappy imager with poor lighting and focus to provide images that can possibly be discriminated effectively across, say, hundreds of thousands of target patterns? What is the false accept/false reject rate for this technology?
Promises and smoke and mirrors are all fine, but why imagine this technology might ever work as advertised?
There are a few companies that have image recognition technology. Some have publicy introduced applications for mobile marketing and some are still working on strategic partners/platforms.
I had to stop at mile 10 on the run. Lungs aren't there just yet.
I saw how long it took for one movie poster to be included in the database...seconds (not from Neven Vision). The image was photographed, using a camera phone in a number of distorted ways and each time connected to content.
From what I saw at CTIA, this technology is already here, and being implemented by certain mobile marketing companies.
Yes, I'd expect that it might not be difficult to add another template into the database IF that database is relatively unpopulated.
But what happens when the database contains already, say, tens of thousands of target patterns -- exactly what you would expect and demand if the technology on a large scale is a success. How well can it discriminate one pattern from another?
You see, THAT is where machine vision, pattern recognition technologies ALWAYS break down. That's when the false accept/false reject numbers become intolerable. For example, that's what has made facial recognition to spot terrorists totally useless -- on the scale they need to be employed, they are so unreliable they are a waste of time at best.
Look, I don't even know of a good OCR program for camera phones for reading URLs -- and that is an infinitely less difficult problem.
Yeah, it'd be great if the technology worked. But why believe that it ever will?
Anonymous, email me imagerecmobileguy@hotmail.com and I'll explain how it scales for a competitor... (please also tell me your background/where you work)
A recent campaign using a movie poster logo.
8 comments:
Scott
What differences are there between Nevenvision and Mobot. Did you get a chance to compare both at the show? By the way how did Ironman go for you?
Larry
The movie poster example simply reinforces a point I made in a post elsewhere on this blog: how well can this technology possibly scale up?
How many movie posters are there? How many ads in ad campaigns? How many company logos? Tens of thousands? Hundreds of thousands?
How do you get a crappy imager with poor lighting and focus to provide images that can possibly be discriminated effectively across, say, hundreds of thousands of target patterns? What is the false accept/false reject rate for this technology?
Promises and smoke and mirrors are all fine, but why imagine this technology might ever work as advertised?
Magic doesn't exist, you know, only technology.
Larry,
There are a few companies that have image recognition technology. Some have publicy introduced applications for mobile marketing and some are still working on strategic partners/platforms.
I had to stop at mile 10 on the run. Lungs aren't there just yet.
Scott, i am an avid reader of your esteemed blog. Just wondering if you had a chance to drop by at NEOMEDIA's booth. Can you tell something about it.
Thanks in Advance
"how well can this technology possibly scale up?"
I saw how long it took for one movie poster to be included in the database...seconds (not from Neven Vision). The image was photographed, using a camera phone in a number of distorted ways and each time connected to content.
From what I saw at CTIA, this technology is already here, and being implemented by certain mobile marketing companies.
I'll have more on this shortly.
I visited almost all of the PWC companies at CTIA.
Some I took pics of their apps, some I took notes and some I just conversed with mgmt.
I am in process of putting together the CTIA summary now.
Scott, your post misses the point I was making.
Yes, I'd expect that it might not be difficult to add another template into the database IF that database is relatively unpopulated.
But what happens when the database contains already, say, tens of thousands of target patterns -- exactly what you would expect and demand if the technology on a large scale is a success. How well can it discriminate one pattern from another?
You see, THAT is where machine vision, pattern recognition technologies ALWAYS break down. That's when the false accept/false reject numbers become intolerable. For example, that's what has made facial recognition to spot terrorists totally useless -- on the scale they need to be employed, they are so unreliable they are a waste of time at best.
Look, I don't even know of a good OCR program for camera phones for reading URLs -- and that is an infinitely less difficult problem.
Yeah, it'd be great if the technology worked. But why believe that it ever will?
Anonymous, email me imagerecmobileguy@hotmail.com and I'll explain how it scales for a competitor... (please also tell me your background/where you work)
Thanks!
Rick
Post a Comment