Enhanced Signature Verification Through the Integration of Diverse Models, Optimizers, and Classifiers in Deep Learning

Different types of documents, such as financial, commercial, security controls, certificate and judicial documents, require signatures for verification. However conventional signature verification processes have difficulties with image-based verification in the face of rising volumes of documents and digital transactions. The transition to digital platforms requires effective verification processes that can cater to real-world complexities. The current study uses machine learning and deep learning techniques, such as (SVM) classification and ResNet-50 for feature extraction, to address these issues. By reducing intra-class variability and increasing inter-class separability, Triplet Loss Optimization optimizes the model’s capacity to learn discriminative features. Moreover, Error Level Analysis (ELA) is used to Determine pixel inconsistencies and enhance feature. Adam Optimizer is used for model optimization to speed up convergence and increase stability while training. Hyperparameter tuning is also applied to improve SVM performance. The efficiency of the model is assessed based on the ICDAR-2011 dataset to validate the strength of this hybrid strategy. Results show that combining Triplet Loss and ELA preprocessing with ResNet-50 embeddings for SVM classification improves signature verification accuracy significantly. In this study, the pretrained model is used to analyze offline signature verification. The external type I is associated with errors (????), where a false signature is accidentally accepted. On the other hand, FRR corresponds to type II error (????), where a real signature error is rejected as forged. The proposed architectural function uses the ResNet-50 for extraction and a Siamese network for classification. This combination gets high accuracy of 98.81%.