<![CDATA[The problem in this research is that the manual signature verification process is still widely used. However, this method is prone to human error and is highly subjective, so its accuracy in distinguishing genuine and fake signatures is not optimal. The pattern recognition extraction process in signatures uses the Alexnet algorithm. This study uses a digital signature image dataset consisting of two classes, with 90 images per class. Furthermore, the signature pattern recognition extraction process based on digital images can be performed using the Alexnet model. The purpose of this paper is to help classify signature types, which can facilitate the medical treatment process. The analysis uses deep learning with Python tools. Explicitly, the total sample size in Figure "Distribution of Classes in Training, Validation, and Testing Data" (image_98f1fc.png) shows that the number of samples for the 'full_forg' class is fewer than for the 'full_org' class. Although the model performs very well on the minority class, the presence of perfect recall for the 'full_org' class will be interesting to observe.]]>
Copyrights © 2025
