Breaking the 4Chan CAPTCHA

A machine learning project successfully developed a model to solve 4Chan's CAPTCHA challenges with over 90% accuracy. The process involved scraping CAPTCHAs, generating synthetic data, and training an LSTM CNN model using TensorFlow and Keras. The final model was converted for use in the browser with TensorFlow.js. Key challenges included data collection, human solver inaccuracy, and compatibility issues between different versions of TensorFlow and Keras. The project demonstrates the potential for AI to outperform humans in certain visual recognition tasks.