Offline Based Signature Verification Using Deep Neural Networks

Every individual has different and unique signatures which are used for authentication and identification of important document. Mostly used to certify the originality of cheques, draughts, certificates, approvals, letters, and other legal documents. A signature is employed in such crucial processes that it is important to confirm its legitimacy. In a number of financial, legal, and professional situations, this kind of verification is essential for preventing document fraud and falsification. In the past, signatures were manually checked against replicas of real signatures. This approach might not be adequate as technology develops and new methods for signature forgery and falsification emerge. Hence, a new effective tool is required to address this issue by adopting a convolution neural network technique. The proposed system assists in determining whether the user's new signature matches the original signature in the dataset. For authentication of offline signatures our research presents an application, which uses deep neural network models like Inception v3, Xception and ResNet50. ResNet50 has outperformed the other two models.