harvardnlp / im2markup

Neural model for converting Image-to-Markup (by Yuntian Deng yuntiandeng.com)
https://im2markup.yuntiandeng.com
MIT License
1.19k stars 214 forks source link

Could you upload the code to generate handwriting formulas? #11

Open ZizhenWang opened 6 years ago

ZizhenWang commented 6 years ago

Hi~I want to generate some handwriting image, can you share your code?

da03 commented 6 years ago

Hmm sorry we have not cleaned up the code yet, but the synthetic handwritten dataset can be found here: http://lstm.seas.harvard.edu/latex/data/ (at the bottom of the page)

TyrionChou commented 5 years ago

Hi,author I want to know you how to replace individual symbols? Getting character position information through pdflatex or writing program to render by yourself

mikewlange commented 5 years ago

Hi. https://github.com/search?q=CROHME. And https://github.com/ThomasLech/CROHME_extractor in particular will help you create the data you need. (I think, of course I don't know the specifics of what you're looking for)

However, just to thorw it out there... most AI/ML is munging data for your models. I highly recommend you incorporate building your own data sets into your education regimen, it will serve you better than pretty much any other data science skill(imho and from years of doing so)

Cheers!

w32zhong commented 4 years ago

@da03 I am wondering how you can replace the horizontal fraction line into hand-written style as shown in your paper, any idea?

da03 commented 4 years ago

Hi @t-k- , it's the same for fraction line or other symbols. I was using mathjax to render latex equations to SVG format, and since in SVG we have paths for each symbolid, we can simply replace each path to the corresponding handwritten path which is kindly provided by the detexify author (but we need to figure out the mapping from symbol to id, which can be found in mathjax code).

w32zhong commented 4 years ago

@da03 Thank you for your reply, that helps a lot.