Back to The Blog


Introduction

Recently, I decided to get back into reading. The last time I can remember starting and finishing a book in a reasonable amount of time was when I was a kid. My main struggle with reading is finding time to do it! I know that’s not a good excuse. After all, I find time for TV every day.

Then I saw an ad for the XTEINK X4, an e-reader that can fit in your pocket!

I envisioned a future in which I would carry this e-reader around and read in my downtime. I could reach for this instead of my phone! But there was only one problem: this e-reader can only read “Digital Rights Management” (DRM)-free e-books. That means I can’t download Libby or Kindle e-books to read. I would need to find DRM-free e-books—which are usually really old, or books whose publishers actively decided not to implement DRM—or I would have to create my own.

After talking with a copyright expert, I became determined to scan my personal book library onto this e-reader. If I physically own the books, create the DRM-free EPUB files myself, and don’t distribute them in any way, then I would be legally in the clear.

Once again, there was only one problem: I had no idea how to do this!

Scanning Books to PDFs

Book scanning is an expensive hobby! There are plenty of purchasable solutions that will run you $1,000 or more. I felt discouraged seeing these prices since I wanted to minimize costs in this project.

Then I stumbled upon a tutorial about a homemade solution. I figured I could handle building that!

I got to work cutting up a cardboard box for a base. This would hold the book in position so I could scan the pages. The tutorial used a camera on a tripod to scan the pages. In my implementation, I used the standard scanning feature on my iPhone.

Book scanning setup

The cardboard book holder.

A quick trip to Hobby Lobby resulted in a pane of clear glass meant for stained-glass projects and a glass cutter. I split the glass pane in half and 3D-printed angled holders to hold each pane in place. This would hold the pages flat so I could scan them clearly.

Completed book scanning setup

The completed book-scanning setup.

Now I was ready to start scanning my first book!

Scanning The Hitchhiker’s Guide to the Galaxy took about two hours. Now that I had a 200+ page PDF, I needed to find some way to extract the words from the pages.

Learning OCR

Thankfully, extracting words from PDF pages is a well-trodden path! Optical Character Recognition (OCR) is the process by which companies like Google, Microsoft, and others make PDFs searchable and copyable.

I worked with AI to code an app that I could provide my PDF to and have the words extracted. However, page numbers and the occasional book or chapter title at the top of pages are common OCR hiccups. I didn’t want to include those in my final EPUB file.

I worked through several iterations of automated ways to ignore these extra page components. None of them worked. Well, that’s not entirely fair. They worked… just not very well!

Eventually, I decided to ditch the “automatic” aspect of the requirement and reworked my app to include a page-cropping feature. Code would split the PDF into image files, from which I could crop out just the text I wanted to OCR. Once I had the cropped text images, I could pass them through my OCR process.

This worked amazingly well!

Of course, there were still some common issues, like particles of dust being interpreted as punctuation or page bleed-through being extracted as Chinese characters or nonsense text. I started using black construction paper behind each scanned page to cancel out the bleed-through, and I used Calibre, an e-book management software, to locate and remove misidentified characters.

After all of this, I finally had an EPUB file of The Hitchhiker’s Guide to the Galaxy!

Final EPUB file displayed on the e-reader

The finished EPUB loaded onto my pocket-sized e-reader.

Conclusion

At the time of writing, I have six EPUB files loaded onto my pocket-sized e-reader. I’m still working through reading all of them! I read far more frequently than I used to. Having my current book in my pocket is the convenience I needed to feel inspired to read.

It doesn’t hurt that I feel obligated to read the books now that I’ve put so much effort into scanning, OCR-ing, and properly formatting them!

This project has taught me a lot about project management and the limitations of technology. When I set out to scan books, I figured there would be countless free, off-the-shelf AI solutions that would work perfectly every time. In fact, I originally went down the path of implementing an AI vision model to read each page. That approach was far more costly and time-consuming than using the tried-and-true OCR method.

While AI is revolutionizing many parts of life, it has not yet surpassed the quality of tried-and-true OCR methods for this kind of task.

If you would like to use the code I created in this project to create an EPUB file from a PDF, click here. That’s a starter ZIP file for the Python project I’ve been using to create my EPUB files.

Best of luck on your e-reader journey!