Document scanner + target word recognizer

Document scanner: input photo of your document to get its smooth pdf version. Optionally recognize words in your document and mark them.

Getting started

Scanner

At first, clone the repo:

git clone https://github.com/balan/document-scanner.git

I assume you have numpy installed in your python. Next you need to install computer vision tools:

pip install opencv-python scikit-image

This is all you need to run your document scanner.

Word recognizer

If word recognizer option is also needed, install Tesseract OCR engine:

sudo apt install tesseract-ocr libtesseract-dev
pip install pytesseract

The default language of Tesseract is English, if your document is written in other language install special package at first (lang responds to your language):

sudo apt-get install tesseract-ocr-[lang]

Usage

Simply run python scan.py with proper arguments to get result. Observe the application interface:

usage: document-scanner [-h] [-i IMAGE] [-p PDF] [--binarize] [-w [TARGET_WORDS [TARGET_WORDS ...]]] [-l LANG]

optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
                        path to input image
  -p PDF, --pdf PDF     path to store document in pdf format
  --binarize            save two document versions: original and binarized (stores both by default, only original when false)
  -w [TARGET_WORDS [TARGET_WORDS ...]], --words [TARGET_WORDS [TARGET_WORDS ...]]
                        list of target words to search in document (no words by default)
  -l LANG, --lang LANG  specify language of text if needed (english by default)

Examples

You may find examples of input images and pdf files in /images and /out folders respectively.

Input and otuput with recognized target words:

Observe marked words connected to sign and signature: