A machine learning project focused on creating a model to solve 4Chan CAPTCHAs with high accuracy. The author collected and processed CAPTCHA data, including generating synthetic data, to train a model using TensorFlow and Keras. Challenges included handling various CAPTCHA types and optimizing the model performance. The final model achieved greater than 90% accuracy in solving real CAPTCHAs in the browser.

15m read timeFrom nullpt.rs
Post cover image
Table of contents
IntroductionTerminologyGetting the DataCreating the ModelUsing the Model in TensorFlow.jsReal-World PerformanceConclusion

Sort: