This repository contains the implementation of the adversarial attack using genetic algorithms on the tesseract OCR.

Example

The protection of the information is one of the key priorities, according to the Mark Zuckerberg: .

Let's use an OCR on this image!

.

Using this tool, one can selectively hide the information!

.

But the human still can read the text

.

Usage

usage: main.py [-h] -i IMAGE [-a] [-p POP_SIZE] [-m MUTATION_RATE]

This program allows user to detect and recognize printed text on images, as well as get the
contents of the image as a text

optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
                        Path to the image
  -a, --annotate        Annotate the image
  -p POP_SIZE, --pop_size POP_SIZE
                        Population size
  -m MUTATION_RATE, --mutation_rate MUTATION_RATE
                        Mutation rate