Handwriting Captcha Solver

About

This project is a Handwriting Captcha Solver that uses Convolutional Neural Networks (CNNs) to recognize and decode handwritten captchas automatically. It utilizes Python, Keras, TensorFlow, and OpenCV to preprocess images, train a deep learning model, and predict captcha characters accurately.

Features

• Captcha Image Processing: Uses OpenCV to preprocess captcha images (grayscale conversion, thresholding, segmentation).
• Deep Learning Model: A CNN-based model trained on a dataset of handwritten captchas.
• Prediction and Solving: The trained model (model.h5) is used to predict characters from new captchas.
• Application Interface: A demo application to test the captcha solver.

Project Structure

File/Folder	Description
application_demo/	Contains test images and a demo application for solving captchas.
emoji_dataset/	Additional dataset (if emojis are used in captchas).
test_captchas/	Sample test captchas for evaluating the model.
app.ipynb	Jupyter notebook for running the captcha solver application.
dataset_processing.ipynb	Preprocesses captcha images (segmentation, filtering, etc.).
character_st.txt	Stores character mappings for captcha labels.
main.ipynb	Main script to run the trained model and solve captchas.
model.h5	Pretrained CNN model for captcha recognition.
model_training.ipynb	Trains the CNN model using Keras and TensorFlow.

Installation & Requirements

Dependencies:

• Python 3.x
• TensorFlow
• Keras
• OpenCV
• NumPy
• Matplotlib

Setup:

1. Clone the repository:

   git clone https://github.com/your-repo/handwriting-captcha-solver.git
   cd handwriting-captcha-solver

2. Install dependencies:

   pip install -r requirements.txt

3. Train the model (optional, if you want to retrain):

   jupyter notebook

   Open model_training.ipynb and run the cells.
4. Run the application:

   jupyter notebook

   Open app.ipynb and follow the instructions.

How It Works

1. Preprocessing: OpenCV processes captcha images (resizing, noise reduction, character segmentation).
2. Training: A CNN model learns character patterns from the dataset.
3. Prediction: The trained model predicts characters in a new captcha image.
4. Application: Users can input an image, and the model will return the solved captcha.

Future Enhancements

• Improve character segmentation for better accuracy.
• Deploy as a web API for real-time captcha solving.
• Support for multi-font and colored captchas.

Contributors

• Your Name (Your GitHub/Email)

License

This project is licensed under the MIT License. Feel free to use and modify!