/CaptchaCrack

Solving Golestan-Captchas by implementing Machine Learning Algorithms.

Primary LanguageJupyter NotebookMIT LicenseMIT

About project

This project aims to read and Solve captchas from the Golestan website. Steps:

1) Collect Captchas:

  • Collect captchas using an API.

2) Make captchas ready for the ML process:

  • Convert captcha backgrounds to black and white.
  • Split the created picture into alphabet and digits by cropping the picture.

3) Implement Classification-Algorithms to make cropped pictures to text:

  • Train the handmade Dataset from Golestan captchas
  • Implement KNN to the trained dataset.
  • Save the model to the .sav file to decrease the program's running time.
  • Generate the captcha's text with 96% accuracy.

How to run?

Install the dependencies.

pip3 install -r requirements.txt

Train Dataset:

Run modelTrainer.py for train model using Datasets.

Solve captcha:

After the model is created and saved in the directory you should run captchaSolver.py to solve the specific captcha.

Files and directories