Starting last year the Android Recent Apps screen allows you to select text from the carousel of screenshots of apps. The way they do this is not by actually running the app, but by doing OCR on the screenshots, so you can actually select text that is otherwise not selectable like text in images by opening up recent apps.
It's acually a feature, you can turn it off in the Pixel Launcher settings/Suggestions/Overview selection.
On Android Pie, it can also be used to share/grab images that you couldn't normally download.
But having grown up in the home of a chemistry professor, I'm having a hard time getting comfortable with this one.
Unfortunately I only noticed the typo after purchasing the domain name and decided to run with it :)
I honestly thought that was the reasoning when I saw the name & the tool. You purify text from images with naptha!
Project Pepercorn alone is enough to get Business Insider writing about it all the time.
from what I gather here, it is an accidental misspelling.
Curiously, if you remove the other 'h' instead, as in 'naphta' (nafta), you end up with the word as actually used in many parts of the world.
Is there an OCR that’s in the ballpark as Google’s image recognition API, and that runs locally? I need to read text from photographed images that I am legally not allowed to upload to google, and Tesseract is virtuallly useless for this.
That's exactly what pressing "Disable Lookup" does.
Here's an online notebook for trying it: https://observablehq.com/@tmcw/tesseract-js-v2-alpha
edit: I just saw that the page says my browser is not supported but I could use more than half of the examples
Stop of doing this !
This stuff is far superior for Chinese character recognition. Does amazingly well with text on background images as long as there is some contrast. Sometimes image cap doesn't work in Firefox, but Chrome derivatives are fine (I'm using with Vivaldi).
"Our Copyfish extension was stolen and adware-infested": https://news.ycombinator.com/item?id=14888010
Google Cloud Vision OCR is incredible for Chinese character recognition. I think it's the same backend as Google Translate uses.
The copy is mostly referring to the text detection algorithm that was being used (It's built on a tweaked custom implementation of Microsoft Research's Stroke Width Transform algorithm)— which was state of the art a few years ago (when I first wrote the words on the website).
Nowadays neural approaches perform a bit better, so I should probably change that.
<[ TEXT RECOGNITION IN PROGRESS / MORE INFO: http://projectnaptha.com/process/ (IDX:1:4&!&!&!&!:XDI) / ELAPSED 23.52SEC / DATE Mon, 09 Sep 2019 16:56:50 GMT / TEXT RECOGNITION IN PROGRESS ]>
Reading untranslated manga would be so-much-easier if it does!
Contact me if you want to chat more about language learning and comics!