kba / awesome-ocr Goto Github PK

euske/pdfminer#138

https://www-i6.informatik.rwth-aachen.de/rwth-ocr/

https://blog.altoros.com/optical-character-recognition-using-one-shot-learning-rnn-and-tensorflow.html

HT @cneud via Twitter

https://github.com/tmbdev/tfndlstm

There are some nice python scripts for various image pre-processing tasks described at https://mzucker.github.io/ eg "Unprojecting text with ellipses", "Compressing and enhancing hand-written notes", and "Page dewarping". The blog posts contain links to the scripts themselves in the author's GitHub repository.

keras-team/keras#3436

via tmbdev/clstm#115 (comment)

https://github.com/Halfish/lstm-ctc-ocr

Written by Ray Smith and other googlers.

https://arxiv.org/abs/1702.03970
http://rrc.cvc.uab.es/?ch=6
https://github.com/tensorflow/models/tree/master/street

Fascinating stuff: http://www.nature.com/ncomms/2016/160909/ncomms12665/full/ncomms12665.html

https://blogs.dropbox.com/tech/2017/04/creating-a-modern-ocr-pipeline-using-computer-vision-and-deep-learning/

I suggest to add https://github.com/tokee/quack by @tokee to the list.

Lios is a free and open source software for converting print in to text using either scanner or a camera, It can also produce text out of scanned images from other sources such as Pdf, Image, Folder containing Images or screenshot. Program is given total accessibility for visually impaired. Lios is written in python3, and we release it under GPL-3 license. There are great many possibilities for this program, Feedback is the key to it, Expecting your feedback.

Active development, looks really cool, screenshots: https://sourceforge.net/projects/lios/

has some facilities for training tesseract.

Source code: https://gitlab.com/Nalin-x-Linux/lios-3

https://www.mindmeister.com/192257150/ocropus-overview-published

https://www.youtube.com/watch?v=gcjCiS9pJ3A

https://www.nr.no/~eikvil/OCR.pdf

Recently, the use of neural networks to recognize characters (and other types of patterns)
has resurfaced. Considering a back-propagation network, this network is composed of
several layers of interconnected elements. A feature vector enters the network at the input
layer. Each element of the layer computes a weighted sum of its input and transforms it
into an output by a nonlinear function. During training the weights at each connection are
adjusted until a desired output is obtained. A problem of neural networks in OCR may be
their limited predictability and generality, while an advantage is their adaptive nature.

In Future improvements

Integration of segmentation and contextual analysis can improve recognition of joined
and split characters. Also, higher level contextual analysis which look at the semantics of
entire sentences may be useful. Generally there is a potential in using context to a larger
extent than what is done today. In addition, combinations of multiple independent feature
sets and classifiers, where the weakness of one method is compensated by the strength of
another, may improve the recognition of individual characters

And it seems that OCR for printed types has been solved for a solid 23 years 👍

The frontiers of research within character recognition have now moved towards the rec-
ognition of cursive script, that is handwritten connected or calligraphic characters.

https://www.youtube.com/playlist?list=PLHgYXaZMqUrjPv7TjFoI1iLG_dYsZznTT

https://digilex.hypotheses.org/153?lang=en_GB

https://github.com/tmbdev/ocropy/wiki/Publications
Are they awesome enough? 😆

ocropus - OCR engine based on CLSTM, Apache 2.0

It's the other way around. 'ocropy' was out before 'clstm' :-)

Maybe the intention was 'LSTM' and not 'CLSTM'.

Using 'ocropus' and linking to ocropy might confuse anyone that does not know ocropy's history.

https://sbdevel.wordpress.com/2016/11/15/prototype-demo-for-ocr-postfix-in-danish-newspapers/

Mentioned in #13, cont'd here:

The synth 90k text recognition dataset is probably a useful addition to the list.

Also of interest: https://github.com/edward-zhu/umaru

@Cloudera Engineering Blog

Also, these mysterious Docker containers by gb17@dockerhub: ubuntu-kafka-ocr and ubuntu-kafka-abby.

hocr-inkscape integration?

https://comsys.informatik.uni-kiel.de/lang/de/res/practical-expercience-with-ocropus-model-training/

http://lib.psnc.pl/dlibra/doccontent?id=358

PDF

http://www.ocrgrid.org/ocrdev.html

Initially made for sanitizing whiteboard photos, but it should be applicable to other images:

https://gist.github.com/lelandbatey/8677901

https://github.com/chreul/larex by @chreul

It would be tremendously useful to have a variety of real-life OCR results and ground-truth.

@suub has some here https://github.com/suub/ocr-engine-results/tree/master/demo/ground-truth

Banti Telugu OCR: https://github.com/TeluguOCR/banti_telugu_ocr
"This framework relies on the ability of a segmentation algorithm to break the text in to glyphs."

Chamanti OCR: https://github.com/rakeshvar/chamanti_ocr
"It will not rely on segmentation algorithms (at the glyph level), making it ideal for highly agglutinative scripts like Arabic, Devanagari etc. We will be starting with Telugu however."

It is hard to guess for me, how good the recognition work, because I don't understand Telugu and I haven't found any results, discussions, blog posts (which I can read). But the project looks IMO very promising from a technical point. CC @rakeshvar

Plumb a PDF for detailed information about each char, rectangle, line, et cetera — and easily extract text and tables.

Seems really useful to reverse engineer layout information.

https://github.com/jsfenfen/whatwordwhere

LITIS/IMPACT Workshop : Recent Development in OCR for Digital Libraries

https://github.com/urieli/jochre

Jochre is an OCR package based on supervised machine learning techniques. It has been applied to several languages, including Yiddish, Occitan and Alsacien.

http://cistern.cis.lmu.de/ocrocis/
https://github.com/kaumanns/ocrocis

https://github.com/michal-h21/hocrtex

This hasn't been updated in a while but it's a really cool approach to creating PDFs from OCR data.

https://github.com/jbaiter/archiscribe

https://arxiv.org/abs/1703.09550

via mittagessen/kraken#36 (comment)

https://github.com/jwilk/ocrodjvu

https://github.com/adgaudio/OcrMatch

http://arxiv.org/abs/1603.03101