![]() | Only 14 pages are availabe for public view |
Abstract Signature verification, a fundamental aspect of identity authentication systems, plays a pivotal role in various domains, including financial transactions, access control, and document validation. This study introduces an advanced framework that leverages Deep Learning Models (DLMs) to revolutionize signature verification techniques. The research highlights the pressing need for continuous exploration and progress in automated signature authentication.In the thesis, five pre-trained DLMs: ResNet50, DenseNet121, MobileNetV3, InceptionV3, and VGG16 are utilized. The proposed models are perfumed and evaluated on four distinct signature datasets: CEDAR, BH-Sig260 Bengali, BHSig260 Hindi, and ICDAR 2011(Dutch).The findings reveal compelling results across these datasets. It is observed that the InceptionV3 DLM based on the ICDAR 2011 (Dutch) achieved the best performance of 100% accuracy, 100% AUC and 100% sensitivity. Moreover, CEDAR Dataset achieves performance with an accuracy of 99.76%, an AUC of 99.94%, sensitivity at 99.76%, precision at 99.67%, and an F1-score of 99.71%. Furthermore, this model demonstrates exceptional computational efficiency, with an inference time of just 13.627 s. |