An Offline Signature Verification Using Convolutional Neural Networks

  • Dewi Suryani , Michael Reynaldo Phangtriastu, and Yudy Purnama

Abstract

Signature is a common tool to determine a person's identity, especially in terms of ratifying confidential documents. This leads to the need for more research on the verification of signatures, in order to prevent misuse of signatures by irresponsible parties. In this study, we focus on offline signatures verification using the Convolutional Neural Network (CNN) Algorithm. The implementation uses TensorFlow and Keras libraries and works on GPUs to perform all experiments. We setup several combinations of training and testing dataset. We managed to achieve 97.71% of accuracy and this result outperformed the previous work. In addition, the number of datasets used as training data greatly influences the accuracy of the CNN model.

Published
2020-05-01
How to Cite
Dewi Suryani , Michael Reynaldo Phangtriastu, and Yudy Purnama. (2020). An Offline Signature Verification Using Convolutional Neural Networks. International Journal of Advanced Science and Technology, 29(06), 4764 - 4773. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/19394