Recognition of Multiple CAPTCHA Types: Using CNN and Combined Dataset
Barış Berk Şengül, Muhammed Burak Alver · 2025
In this study, a CNN-based model capable of solving multiple CAPTCHA types was developed. The model was trained using both real and synthetic CAPTCHA datasets to enhance generalization. Initially, the model was trained separately on the Kaggle dataset and a synthetic dataset generated using the Python captcha library, and their performances were analyzed. Subsequently, both datasets were combined to evaluate the model’s ability to solve diverse CAPTCHA types. Experimental results indicate that the model trained with the combined da-taset achieved the highest accuracy and improved generalization compared to models trained on a single dataset. Future work will focus on enhancing the model’s capability to handle more complex CAPTCHA types and integrating it into real-time applications.