A convolutional neural network model built with Keras for recognizing captcha images generated by the Laravel Captcha library.
captcha_cracker is a convolutional neural network model built with Keras that recognizes and cracks captcha images. It specifically targets captchas generated by the Laravel Captcha library, providing a machine learning solution to automate captcha recognition. The project includes a complete pipeline from dataset generation to model training and evaluation.
Developers and researchers interested in machine learning, computer vision, or security testing who want to understand or implement captcha recognition systems using CNNs.
It offers a simple, reproducible implementation focused on a specific captcha library (Laravel Captcha), with pre-trained weights and clear training scripts, making it accessible for educational or experimental purposes.
卷积神经网络实现的验证码识别
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Trained exclusively on Laravel Captcha images, making it highly effective for that library, as demonstrated by the online demo and focused dataset generation.
Includes scripts for generating datasets, preprocessing images into RGB matrices, and model training, offering a reproducible example for educational purposes.
Model weights are saved in weights.hdf5, allowing immediate use without retraining for quick testing or demonstration.
Built with Keras and clear code, providing an accessible entry point for learning convolutional neural networks applied to captcha recognition.
Only works reliably with Laravel Captcha-generated images; it lacks versatility for other captcha styles without significant retraining efforts.
Requires Ubuntu 16.04 and Python 3.5.2, which are old and may not be compatible with modern systems, leading to setup challenges.
Trained on only 2000 captcha images per group, which risks overfitting and poor performance on unseen variations or more complex captchas.