No OP, but I'm aware of a few:

- scandit: https://www.scandit.com/developers/

- cognex: https://cmbdn.cognex.com/

- aspose: https://products.aspose.com/barcode/

In open source, I heard of libdmtx[0], but not much happens on it these days. I also saw html5-qrcode[1]

[0]: https://github.com/dmtx

[1]: https://github.com/mebjas/html5-qrcode, which directly depends on a port of zxing to javascript.

Fascinating, thanks a lot for these.

Scandit and Cognex have iOS apps, and Aspose has a web recognizer demo.

I tried all three with a 2D QR code (my vax card!) from which I truncated ~2-3 bit-"rows" from the bottom. None of the apps/site detected the code.

This is truly not a thorough test -- the 2D rows may be more information-dense than I think i.e. they are not recoverable using the built-in error correction (basically I'm removing too much information -- the equivalent of a long/deep scratch on a CD).

--

I'm just trying to get an intuitive sense of how much damage a QR code can sustain and ... I guess if there's such a thing as QR code forensics (there must be...??) for decoding partial codes versus not detecting "anything".

Thanks again for the links. Super interesting space

When generating a QR code, you can decide how much redundancy to include, making the QR larger for the same data but more resilient to "scratches".

Thanks. Looking into it [0], the QR code I'm testing uses the "up to 7% damage" error correction level L, which makes sense since there's already a lot of data being packed in the (vax proof) code.

... I still want to find a debugging/verbose QR decoder :-)

[0] https://blog.qrstuff.com/2011/12/14/qr-code-error-correction

There are a lot more QR code libraries available, and while they're conceptually very similar, the open source DataMatrix libraries are much less developed and are far less reliable.

I tried libdmtx, but it was extremely slow, and also it had problems with most of my barcodes, even after I cleaned up the images as much as I could.

We ended up using one from a company called 2D technology group. They were the only one we found that supported ARM Linux and did grading of codes.