Captcha Recognition Using CNN

Abstract

Captcha (Completely Automated Public
Turing test to tell Computers and Humans Apart) is a
widely used security mechanism to prevent bots and
automated systems from abusing web services. This
project proposes a deep learning-based approach for
recognizing Captcha text using Convolutional Neural
Networks (CNN). The system involves preprocessing a
dataset of Captcha images, training a CNN model to
recognize character sequences, and validating the
model's performance with test images. By converting
images to grayscale and normalizing pixel values, the
model is trained efficiently and achieves high accuracy.
The system demonstrates that CNNs are highly effective
for image-based text recognition, achieving a loss of
0.033, equivalent to an accuracy of approximately
99.967%.