Skip to main content


I co-authored a Mozilla blog post about our #a11y work in celebration of International Day of Persons with Disabilities. I'm posting it a bit late here, but better late than never, I guess. :)
How we’re making Firefox accessible and delightful for everyone: blog.mozilla.org/en/products/f…
#a11y

Peter Vágner reshared this.

in reply to Jamie Teh

Is the image recognition feature exclusive to the MAC or is it behind a feature flag for other platforms as well?
in reply to Peter Vágner

@pvagner Currently Mac only. Windows will come next. The reason is that it depends on the OCR facilities provided by the operating system.
in reply to Jamie Teh

For the future would it be doable to wire tesseract into this so we can get a bit of joy on linux as well? Or is it too complicated?
Still I believe it's a nice design choice it does not use its own cloud implementation.
in reply to Peter Vágner

@pvagner I believe there has been some discussion there. Even though my team has been helping to ensure the UI is accessible, I'm not on the actual team responsible for this feature. I'll see what I can learn though. CC @asadotzler
Unknown parent

Jamie Teh
@asadotzler @pvagner Oh, there are absolutely performance and accuracy concerns with Tesseract. OS OCR is far better on Windows and Mac. However, Tesseract might be the only option if we're ever going to support this on Linux.
Unknown parent

Jamie Teh
@asadotzler @pvagner That's certainly a concern, though I think it's less of a problem with OCR. I think OCR has been around long enough that there's some intuitive understanding that it isn't perfect, plus when it fails, it (usually) fails in such a way as to be very clearly incorrect, rather than plausible enough to be misleading. That said, there's also the question of the technical feasibility of bundling Tesseract for a single OS, etc.
Unknown parent

Jamie Teh
@asadotzler @pvagner Ah, intriguing. I'm not sure.
in reply to Jamie Teh

@Jamie Teh @Asa Dotzler Since tesseract v4 accuracy is greatly improved. For adding text layer to scanned pdf files I am using an app called #ocrmypdf which uses #tesseract and it works sufficiently well for me. There are even some unofficial builds of tesseract for #android. Other than self packaged tesseract there is mlkit by google people are using on android for recognizing images. Performance and interfacing with tesseract may not be that great, I understand.
in reply to Jamie Teh

so why mozilla took so long to fix no heading issue in macOS 13?
in reply to pax

@pax We have a small team and a lot of high priority work to juggle. When that bug was first filed, MacOS 13 was in beta. None of our team were able to run it; running a beta OS (or even a newly released OS) is not without its risks. That bug is also very obscure; it seems to be due to a somewhat inexplicable change in VoiceOver, and as usual, Apple's documentation is poor to nonexistent.
@pax