tessdata VS virtualpaper

Appears to be a nice wrapper around Tesseract:

https://github.com/tesseract-ocr/tessdata

https://en.wikipedia.org/wiki/Tesseract_(software)

The demo of course works perfectly on a Mac as this is already built into Ventura.

  In November 2020, Brewster Kahle from the Internet Archive praised Tesseract saying:

The language packas are available at: https://github.com/tesseract-ocr/tessdata

Then you could try different engines. --oem 0 disables the shiny new neural network stuff and uses the classic Tesseract engine; --oem 1 does the opposite. Try both, see which works best in terms of performance and accuracy for your particular use-case. You'll need to have training data for the legacy engine, though. These would work, or you could try tessdata_fast, which is specifically built for speed.

Go to the GitHub repo and get the trained data of your desired language --> https://github.com/tesseract-ocr/tessdata

Capture2Text uses Google's Tesseract to perform OCR, so the accuracy when detecting foreign languages depends on the quality of the data files. Ones provided by Tesseract itself can be found here.

PGS graphical subtitle rendering is not supported in web browser. But we can convert it into text subtitle by using ML assisted OCR. https://github.com/tesseract-ocr/tessdata

FROM ubuntu:18.04 RUN apt-get update RUN apt-get -y install curl RUN apt-get -y install wget unzip RUN mkdir tess5 RUN cd tess5 RUN wget https://github.com/tesseract-ocr/tesseract/archive/refs/tags/5.0.0-alpha-20210401.zip RUN unzip 5.0.0-alpha-20210401.zip RUN rm 5.0.0-alpha-20210401.zip WORKDIR /tesseract-5.0.0-alpha-20210401 RUN apt-get -y install autoconf automake libtool pkg-config libpng-dev libjpeg8-dev libtiff5-dev g++ # or clang++ (presumably) RUN apt-get install zlib1g-dev RUN apt-get -y install libleptonica-dev RUN ./autogen.sh RUN ./configure --enable-debug RUN LDFLAGS="-L/usr/local/lib" CFLAGS="-I/usr/local/include" make RUN make install RUN ldconfig WORKDIR /usr/local/share/tessdata RUN wget https://github.com/tesseract-ocr/tessdata/raw/master/eng.traineddata ENV TESSDATA_PREFIX=/usr/local/share/tessdata/

You'll also need to go and get the trained language data from here: https://github.com/tesseract-ocr/tessdata/blob/master/eng.traineddata

The code looks fine, I think for whatever reason the 'MathEquations' network just does a poor job of recognizing the equations. The support package that includes the language is based on this open-source tessaract repo which seems to struggle with math equation recognition (at least based on this issue).

The software I built for myself: https://github.com/tryffel/virtualpaper. It does not support the folder hierarchy you suggested, instead everything is accessible through key-value metadata and the search experience is really good IMO.