Automatic reading of text in the image of a document

The aim of this repository is to create a program capable of reading a text written in an image. The image contains a photograph of a page of a document which can be taken with a free angle of view (inclination, perspective effect, etc.). It is therefore a question of estimating this angle of view in order to straighten the image so that it seems to be taken from the front. Then, it is necessary to implement a detection of lines, words and then letters in order to reconstitute the entire text.

Prerequisites

• matplotlib
• opencv
• numpy
• scipy
• python=3.8
• jupyter
• pytesseract

See more in requirements.txt and environment.yml

Installation

1. Install Jupyter Notebook by following the instructions from the official Jupyter website.
2. Install the required Python libraries by running the following command in the terminal:

pip install -r requirements.txt

or create a virtual env with conda (preferred method)

conda env create -f environment.yml

3. Download and install Tesseract OCR from the official Tesseract GitHub repository here.

Usage

1. Clone or download the repository to your local machine.
2. Open the auto_read.ipynb file in Jupyter Notebook.
3. Replace the path to the sample image in the code with the path to your own image.
4. Replace the path to the tesseract.exe in the code with the path relative to your own installation.
5. Run the cells in the Jupyter Notebook to see the output.

With CLI

python auto_read.py [path to your image (.png or .jpg)]

Result

The output will be the extracted text from the document image. The text will be displayed in a readable format and can be used for further processing or analysis.