I recently purchased a small refrigerator whiteboard and it's been really amazing with the combination of my iPhone's ability to take a picture of my handwriting (script or cursive) and copy/pasta into a text. It's not always perfect (nor is my handwriting!) but it's good enough to just replace a character or two and hit send.<p>This really tickles a bunch of things for me:<p>1) I am not sending a whole image (it's efficient)<p>2) I don't have to type/swipe at all (I'm not looking at a screen)<p>3) My s/o has easy access to the list at all times (it's not in a cloud)<p>4) It requires no power to update/maintain (markers last a long time)<p>5) It just feels so natural to grab a marker and write on the fridge when I exhaust something in that same fridge.
This is very cool. Here is interesting application of something like this. My handwriting is pretty bad, and worse still when writing fast. When I am teaching, a lot of what I write is worse than I would like it to be.<p>I could teach a system like this my very slow neat handwriting. And then as I write on my whiteboard while teaching, it replaces my quick bad handwriting with the neater handwriting.
From the title, I naively assumed this article would be about people relearning to make legible/beautiful handwritten notes after losing this ability. That is something I’m currently struggling with after many years of too much typing and not as much handwriting.<p>Google’s actual research does help people like me, by making our notes less awful digitally. But I’d love not to be dependent on tech innovations to make my handwriting better.
I tried to use tesseract for OCR, 10 years ago, it recognized English good enough. tesseract was also developed by Google if I am not mistaken, but open source.<p>I tried to use it then, for non English language, for Greek, and it was very bad.<p>Happy to see some good OCR research based on transformers.
What is currently state-of-the-art when it comes to detetcting handwriting from photos?<p>Tracing strokes is nice but I would be more interested in converting my handtaken notes to markdown.
Very interesting experiment. I've been working on a handwriting application [0] for the past couple of years and incorporating the ability to take a picture to convert it into digital ink would be really nice.<p>[0] <a href="https://scrivanolabs.github.io" rel="nofollow">https://scrivanolabs.github.io</a>
I still look forward to have programming environments on tablets that are able to use pen input, instead of forcing us to carry a bluetooth keyboard.<p>Apparently not something that anyone cares as business opportunity, because most likely most folks wouldn't pay for it, sadly.
A model that could turn "offline" handwriting (the ink on the page) into "online" (order and timing of the strokes) I think could be really useful for a historical HTR pipeline... but ultimately, we need end to end.<p>Why is historical HTR so neglected in all multi-task model evaluation benchmarks? There are millions of un-indexed handwritten historical documents which could give us a so much better understanding of our recent past. For that matter, it could give <i>models</i> much better understanding of our recent past.
Can we replace / augment the keyboard with a white (or marked) paper to write on ... that is in view of a camera that has real time OCR<p>Feels like it will be a good addition to input devices
Eric Hebborn (<a href="https://en.m.wikipedia.org/wiki/Eric_Hebborn" rel="nofollow">https://en.m.wikipedia.org/wiki/Eric_Hebborn</a>) wrote "Italico per Italiani" (available in Italian only, alas <a href="https://www.amazon.co.uk/Italico-italiani-moderno-trattato-calligrafia/dp/8894272273" rel="nofollow">https://www.amazon.co.uk/Italico-italiani-moderno-trattato-c...</a>) as a textbook/manual to improve day-to-day, practical calligraphy.
Maybe there is something similar in other languages too...
i still write notes daily and already this year have finished 4 notebooks...however, things i need to review are typed, i write to keep things in my mind as they are being discussed...example, if i encountered an error while reviewing a program execution i would write it down "encountered error during attempt to do x" but i would also type it in my notes in vastly more detail with screenshots and other points...handwriting to me is almost like tagging it in my mind so i just don't forget that it happened.
Its not the writing PART!<p>It is the non automation part!<p>The experiment you can do to verify my point:<p>1. Write code by hand in plain text editor.....<p>What you notice....programming syntax and knowledge becomes easier to retain and re-learn.<p>No joke I do this several times a week.
As a lay reader, it's fascinating to see how much oomph we're getting out of LLMs on non-language-related tasks by figuring out clever encodings/linearizations.
This is very interesting. I had this idea of imitating human handwriting in my bucket of todos for machine learning models, but never got to it. I guess we aren't far from it.
seems overkill... for a bunch of reasons tbh. I guess the golden sledgehammer is just too tempting.<p>taking photos of notes is a weird compromise that nobody really wants... we need better tools not better post processing
such an exciting research project! I can imagine the impact this could have on education, e.g. handwriting notes of teachers in digital copies; or even preserve old documents in their digital counterpart
Can I just get good OCR for handwritten text? The last model that claimed to be "the best" was atrocious, only worked on PDFs of papers. ChatGPT is pretty decent but I was hoping for an offline, tailored solution
> We present a model to convert photos of handwriting into a digital format that reproduces component pen strokes, without the need for specialized equipment.<p>Call my a cynic but this feels like a free way for Google to pull more data for training.
Now I'm wondering if they can use a similar architecture to derender paintings. It would the quite something to have a stroke perfect recreation of the Mona Lisa drawn by a modified pen plotter.
Tbh my handwriting is write only. Through decades of computer use it has deteriorated to the point were it is illegible even to myself.<p>Still. I will always be notetaking (and doodling) during meetings as it helps me order my thoughts.