CAPTCHA Recognition Using Convolutional Neural Networks

The abstract of the research paper "Captcha Recognition using CNN" tackles the CAPTCHA recognition problem using CNN. The model was trained over many years, more specifically over 50 epochs, to ensure robust learning. Obviously pretty competent at spotting CAPTCHA images, the model achieved 0.8702 as the last accuracy. The Adam optimizer was used to try to reduce the loss function with a final loss value of 0.327. Carefully chosen to optimize the performance of the model, batch size of 32 and learning rate of 0.001 were basic hyper parameters. This approach offers a constant reaction that might be used in practical environments to improve security systems, therefore resolving the problem of automated CAPTCHA recognition. CNN architecture along with highly calibrated hyper parameters gives outcomes for this effort.