Way cool! I am impressed.
Project Naptha: a browser extension that enables text selection on any image
41–50 of 141 posts
Re: Project Naptha: a browser extension that enables text selection on any image
#42@antimatter15, i have a project that does client-side image analysis and decompses document structures. it looks like your OCR code would be a great replacement for the server-side Tesseract ocr i currently use :) here's what the project does now with js + web workers: http://i.imgur.com/QvXSkY2.png processing time is the code is open source, though using it isnt yet polished. i'm working slowly on a blog post series…
The OCR code is an Emscripten port of the GPL-licensed Ocrad program. I published it on Github a few months ago, http://antimatter15.github.io/ocrad.js/demo.html
But in my experience, the recognition quality isn't good enough to replace Tesseract if you have that capability.
Re: Project Naptha: a browser extension that enables text selection on any image
#43Awesome. How does this affect page performance?
Re: Project Naptha: a browser extension that enables text selection on any image
#44Now the NSA will be reading the contents of your animated GIFs.
The NSA reads it before its animated. What's new?
Re: Project Naptha: a browser extension that enables text selection on any image
#45@antimatter15, i have a project that does client-side image analysis and decompses document structures. it looks like your OCR code would be a great replacement for the server-side Tesseract ocr i currently use :) here's what the project does now with js + web workers: http://i.imgur.com/QvXSkY2.png processing time is the code is open source, though using it isnt yet polished. i'm working slowly on a blog post series…
The OCR code is an Emscripten port of the GPL-licensed Ocrad program. I published it on Github a few months ago, http://antimatter15.github.io/ocrad.js/demo.html But in my experience, the recognition quality isn't good enough to replace Tesseract if you have that capability.
it would be very useful to maybe just use part of the code. (the part that detects where there is text, rather than what the text is)
Re: Project Naptha: a browser extension that enables text selection on any image
#46Very slick!
Does it automatically start OCRing every image, or does it wait for a user to try to select the image text?
Asking because I'm concerned about this decreasing performance.
Re: Project Naptha: a browser extension that enables text selection on any image
#47Very slick! Does it automatically start OCRing every image, or does it wait for a user to try to select the image text? Asking because I'm concerned about this decreasing performance.
It waits until you start selecting the image text, but the text detection starts when your cursor moves toward an image. It uses WebWorkers extensively, so on a multicore system, the performance shouldn't be hit. I haven't noticed an effect on battery life, but that's not out of the question.
Re: Project Naptha: a browser extension that enables text selection on any image
#48Holy crap, antimatter15 does so many cool things. I keep finding things that are really cool and then scroll down to find they are all written by him. First Shinytouch, then Protobowl years later and now this. And he's only a year older than me (19) so it isn't that he's had more time. Check out his Github profile for more of his projects: http://github.com/antimatter15
Re: Project Naptha: a browser extension that enables text selection on any image
#49I would like it if this would work on ANY text in webpages.
Too many webpages make it too hard to select even actual plain text.
Re: Project Naptha: a browser extension that enables text selection on any image
#50Certainly a cool idea but it didn't work fine on an XKCD comic: http://www.xkcd.com/ bottom line here is recognized as: "T1EN°5'lI'ONAl.1?E£ONNH\56PNCE(YHCEPlP6fiN(N)SURLH’PR3AO-i‘lDlsIr'£7E‘5IJ%z"
Randall Munroe's handwriting is a bit difficult to OCR because a lot of the letters are smushed together close enough that the it's not possible to unambiguously segment the text into distinct letters (which is a necessary first step in any OCR engine that I'm aware of). Maybe Google's (or Vicarious's) magical convolutional neural net that can solve CAPTCHAs would fare better.
> it's not possible to unambiguously segment the text into distinct letters (which is a necessary first step in any OCR engine that I'm aware of)
In my experience, the ability to handle overlapping letters (which is very common on type-written text and professionally typeset material) is one of the key things that separate the relatively lightweight OCRs (like Ocrad and GOCR) from the big complicated ones (Tesseract, Cuneiform, Abbyy etc). Whitespace character segmentation cannot be taken for granted if you want to do any useful OCR of "historical" material.