ldenoue

Why does it feel important to you that the speaker be standing?

I also made one using BodyPix.js see https://www.appblit.com/handoverface

Detecting where tables are is still an active research areas. Once we know where on the page, it’s easier to parse out their structure.

Or maybe use a text corpus to infer likely next words between 2 lines of characters so we can infer the reading order and columns using the text.

That’s the approach I’m using to reformat “reflow” PDFs for mobile in my app https://readerview.app/

PDFs of research papers are what I often read on mobile, and I find it hard. Check out the app I published if you have an iOS device: it’ll reformat your pdf so the content reflows to fit your screen. https://www.appblit.com/pdfreflow

Yes but converting a pdf to ePub is not easy (to make the ePub reflow the text at least). What I’m working on could output as ePub.

And on desktop you’re fine dealing with PDFs in the browser? Just posted a video demo of what a pdf could look like once reflowed as HTML (look at the HTML version after second 10) https://www.youtube.com/watch?v=mSRFcLovatI

thank you. I never considered it would be useful on desktops too. I guess we now have expectations about how web pages behave in a browser, and what we can do with them; but when the browser loads a PDF, all that is gone, or not working the same way.

What kind of analysis and extraction? Figures, tables? Other?

Google App Engine has been great for me and it’s free. I’ve hosted several personal sites on it since its early release for python, but you can now use many languages. I host now https://www.appblit.com

It’s a good concept but limited posts for now. I wonder why we don’t write a twitter search engine that automatically finds these kinds of job postings and aggregates them by location and company.

What is or was your platform? Wouldn’t giving free access to the app be a reward?

No, pdf.js displays PDF documents in the browser, but not as responsive HTML (it actually renders pages into a CANVAS element and then adds text selection by overlaying DIV elements over this canvas)

My app converts the PDF to a truly responsive HTML document, so you can view this document as any screen size (ie there’s no fixed layouts used)

I have another app based on the same tech but that keeps all formatting. It’s called Pdf Reflow on the App Store. But I gather from the community that a pure HTML converter would be greatly appreciated, even if some characters are lost in translation...

How come the text is centered instead of left or justified?

You can also try the web app at https://docushow.com

Mobile version here https://docushow.com/viewdocchris?url=http%3A//www.cs.cmu.ed...

But if the pdf looked readable on your Android then you might like to read them right there as opposed to later on your laptop, right?

That’s why I am doing it: for myself first but I’m still trying to understand who would really use this.

Oops that’s a problem with finding columns. I’ll see how to solve it. Thank you!

What link was causing the crash?

New is the ability to highlight pdf documents (in addition to web pages)

Do you think PDFjs could be extended to write new text inside the PDF, all client side? At the moment I’m only aware of PDFBox from Apache to do this, but that’s server side JAVA...

shameless plug: give https://docushow.com a try!

Thanks for linking to https://docushow.com Also a work in progress, but PDF reflow is a hard problem so you never ship if you want to solve all cases :)

Your solution using the LaTex source generates really nice HTML, congrats!

When I tried the client side version on the same image, it wasn't good compared to the server side. Are you using a different model on the client?

Looking forward to a YOLO or SqueezeDet or SSD+MobileNet version!

Would be great to port YOLO on your library; always an impressive visual demonstration

I taught my 8 years old girl using JavaScript and we made a simple game (numbers falling from the top of the screen and you have to click them to make a number: e.g. "3" and "4" because the number to make was "7").

It eventually became PopMath (which I reimplemented in ObjectiveC for the iPhone) but she understood the concept of loops, random number generation, collision with sides (so x coordinate had to be reversed) and event handling (function to get the user clicks).

I also did the same with my son when he was 9, with another game where the player would need to use arrow keys to navigate a maze.

Also in JavaScript because they can see the results right in the browser, can easily add graphics, and even sound effects.

I think simple games are a great teaching tool.