Hacker News Reader

Comment by grimen | Hacker News Reader

The real thing will be image scanning and nothing else. I'll post a video soon here on HN to prove that images are way better "barcodes" than both 1D and 2D/QR barcodes.

What I can say now is: 1. QR codes don't work on distances - state of the art image algoritms do. 2. QR codes mess with the brand appearance - basically H&M don't want a QR code all over an ad. 3. QR codes fail when there's reflection, dark dirt, and very skewed camera - state of the art image algoritms don't. ...

I wish I could post our video today to prove how much more convenient image scanning is - and what we use it for - for non-believers, but I have to wait a few weeks. Until then check YouTube.

View on HN

They've been doing the image scanning thing a bit in Sweden, where you can take pictures of bus stop ads and be taken to a site. The problem as I see it is that it needs a database connection, which means

* You need a network connection for it to work, so it won't work everywhere (subway ads)

* Since there's a network service involved, it can't be an open standard, and users will need 20 different apps installed for all the different service providers, and figure out just what app to use for this damn ad.

* Due to the network roundtrip, you don't have instant feedback that the scan worked, which is a problem since

* Performance is poor in the dark.

QR codes also solve the problem of telling people that there's something actually there to scan.

1) Network works well in metro for me mostly, though it's a valid point that it might suffer here and there. But metros is just a smal fraction of the possabiities we - and retailers we talked with - see.

2) It won't be 20 apps, becuase there are only a few companies in the world that has sophisticated algorithms enough - many algorithms you can read about on the web are not realiable enough. So far no one has done a consumer product but hundreds of QR-based apps, I would say that this say a bit how hard it is. Anyone can craete a QR app based on some open-source libs.

3) Well, our demos show something else - ~1-2 seconds.

4) Flash. And scenario for scanning stuff in night dark is...not really a descent secenario anyway if it's that dark people have hard to see.

5) That's an assumption based on today's reality, because no viable succeedors are available yet that do more valuable things. Yet I haven't heard of anyone using QR codes more than once or twice for fun even though it been "hot" since for half a decade or more.

Wait, what? Can't all of these magical "state of the art image algorithms" be applied to scanning QR codes?

These algoritms can look for features in the image in a matter of milliseconds and match those against known ones (which is not a problem, the set can be huge withour performance impact if handle correctly). If you want to detect QR codes that works per definition of image detection, but you are not bound to such - a red shirt with a tiny tiny logotype is good enough characteristic.

Not sure if I answered your question though. Did I? :)

But how do you know you can scan it in the first place?

Well, in our case we have the long-term vision of being the first choice for bying products via image feature detection. We believe that people in the future will expect anything to be scannable, in same way anything is searchable. In both cases you might get false positives with a rating how good match it was. Though, our view is that most people haven't seen the state of the art solutions in use for consumer products. We want to change that, very soon.

So. You don't.

Well, if majority of retailers want this - then expect them to make customers use it asap. You got right to be pessimistic about our vision, time will tell if you are right. One thing that I'm sure of though; QR codes will not survive for long.

...well NFC/RFID will be useful for some scenarios - but they don't work on distance reliably if you want to target "something" directly.

re: brand appearance...

http://hackaday.com/2011/08/11/how-to-put-your-logo-in-a-qr-...

Hey man, try to pitch that to H&M or Armani. xD

Reply on news.ycombinator.com