AI-Powered Approach For Verifying Handwritten Signatures Using CNN

Abstract

In this paper, we present a system that uses deep learning approaches to verify offline signatures very accurately.  While many industries rely on signature verification as a biometric authentication approach, automated verification is difficult due to the signature's inherent unpredictability.  To circumvent the drawbacks of conventional approaches that depend on manually created features, this research use Convolutional Neural Networks (CNNs) to autonomously extract unique characteristics from signature photos.  A CNN-based classifier that uses transfer learning with a pre-trained EfficientNetB0 model to improve performance is described in detail in the article.  Methods like Accuracy, F1-Score, ROC-AUC, and Equal Error Rate (EER) are used to assess the system's performance after it preprocesses signature photos and trains the deep learning model.  We want to build a signature verification system that is efficient and trustworthy. It should be able to forecast outcomes in a structured CSV format and function well with administrative, financial, and legal processes.  The purpose of these results is to prove that deep learning is the best solution for solving the problems associated with offline signature verification.