Tesseract.js VS scantailor-advanced

What do you have against tesseract.js?

Are you able to use a OCR library in the browser? Here is one I found

But keep in mind, PDFs can be fully or partially made of images, and if you want text out of those, you will need to OCR them. There's tesseract.js for that.

Tesseract.js

YOu could try using tesseract.js to OCR the images. Though might be a challenge to figure out the times/days of the classes.

I've never used it - but Tesseract.js looks interesting. Perhaps you can use it as a foundation on which to build something like the AWS tool (which seems to work by overlaying an SVG over the top of the image, then manipulates that SVG when user clicks on a results tag to get the word/phrase's highlight box to display)?

I tried researching a little on this topic, so from what I understand to achieve this feature, I would need a library that supports Optical character Recognition (OCR) such as Tesseract.js or Google ML Kit to do it. However, I'm not too sure if these libraries will be supported on React Native apps.

It's definitely one of the best for on-the-fly OCR, but I think a server-based implementation using Google's Vision API would be way better. Definitely not mine, the AI used here is way beyond my skill level - link: https://tesseract.projectnaptha.com/

There's also https://scantailor.org/ (and a maintained fork at https://github.com/4lex4/scantailor-advanced ) which semi-automates unwarping and other corrective tasks in scanned books.

ScanTailor Advanced (downloads are in the sidebar, labeled as "Releases") is a little more complex. It takes a folder of images and performs a series of steps to produce good book scans, though it might be useful for your purpose as well. The first step after importing images is to split them up, and ScanTailor does this by looking for straight lines that might indicate a gap between pages. Now, I don't know if this will handle any more than two pictures per image, but it will at least handle doubles. ScanTailor will also attempt to automatically deskew pictures, and then there are a few steps that you'll need to take to make your pictures come out nicely.

My preferred method is to take pictures of all the pages of the book (Open Camera has a nice option to take a new picture every n seconds), optionally touch them up with ScanTailor (automated), and then turn all the images to a PDF using NAPS2 (which will OCR the text as it goes in).

Next is the cleanup. Scan Tailor is the best game in town for this, but it's a dead project. Instead, there are two forks that have picked up where the original developers left off. Scan Tailor Advanced is my current fork of choice, though Scan Tailor Universal tries to add new usability features. For whatever reason, only Advanced makes full use of my CPU, so it's several times faster than Universal for the time being.

No problem. https://github.com/4lex4/scantailor-advanced/releases

If you are interested in more ways to treat a scan, ScanTailor is definitely an option. I am using a fork called ScanTailor Advanced which is available under GPL-3.0 on GitHub. With this tool you can also crop your images and apply options like threshold or posterization to improve readability while further reducing the file size.

I use a £15 arm with a vice grip for my phone from Amazon, copy the files to my laptop and then run a bash for-loop of the tesseract CLI over the resultant files.

I use https://github.com/4lex4/scantailor-advanced to deskew the images and generate the PDF.

It isn't perfect but my purposes are more around research than publication, so, YMMV!