I tried two QR readers on the photos in the article. They did not detect the code.
Would it have been much more effort for the photographer to take a photo that was scannable? Or is there just too much variance in the arrangement and no reader would actually scan it even without a photo?
I scanned it with WeChat and the public WeChatID of the company popped out - haomengshenghuo, whose direct translation of english is: life with good dreams.
It worked with the scanner in the built-in camera app in iOS 11, but I had to manually adjust the brightness (issue with scanning off a computer screen?)
Might be a problem with the combination of your monitor and your phone camera. I tried Barcode Scanner on Android, and it scanned it without a hitch. Hell, I had more trouble scanning a QR code I hand-drawn in blue on white paper...
An old QR app I had for a long time (from when QR codes were first introduced, many years ago) failed to read it too. Some of the bits look ambiguous too --- e.g. taking the upper left corner bit of the upper left locator square to be (0,0), what's the bit at (16,11)? It's both noticeably darker and lighter than its neighbours, like it was cut out and then grew back slightly.
Those which can't read the code probably use traditional thresholding to reduce the input into 0-1 first, while those which can are using some variation of https://en.wikipedia.org/wiki/Soft-decision_decoding algorithm.
It doesn't help either that the middle was blanked out, eating into the error correction budget.
Yes, my car had one too, so anybody could tip it when they thought I was driving well. Fun Bitcoin project of 2012. Though hardly anybody knew what it was, it actually did get some tips.
On a side note about 9 years ago I was involved in a project to study "digital" camouflage and exploitation of the image recognition systems for a GIS solution primarily for governments where patterns were used to "fool" algorithms that identify objects such as buildings and roads.
We did identify a few issues with an ESRI/IBM solution, I always wondered if these techniques are being actively used in the field.
It won't fool a human analyst but since satellites often operate in mosaic mode these days and generate 10,000s of images per tasked pass CV is often used to identify objects of interest.
16 comments
[ 3.1 ms ] story [ 31.6 ms ] threadWould it have been much more effort for the photographer to take a photo that was scannable? Or is there just too much variance in the arrangement and no reader would actually scan it even without a photo?
http://weixin.qq.com/r/WD-i_pfE3sqhrdOZ92pK
Visiting that link simply forwards me to
http://www.wechat.com/mobile/en/
Rather anticlimactic.
You need to install WeChat and use the app to scan it to go the official account of the tourism office.
Those which can't read the code probably use traditional thresholding to reduce the input into 0-1 first, while those which can are using some variation of https://en.wikipedia.org/wiki/Soft-decision_decoding algorithm.
It doesn't help either that the middle was blanked out, eating into the error correction budget.
So did Google Goggles.
We did identify a few issues with an ESRI/IBM solution, I always wondered if these techniques are being actively used in the field.
It won't fool a human analyst but since satellites often operate in mosaic mode these days and generate 10,000s of images per tasked pass CV is often used to identify objects of interest.
This QR code reminded me of this.