Diploma Verification Through Multi-Modal Image Processing: Face Detection, Perforation Text Recognition, and System Architecture Evaluation

This research presents a novel approach to digitally verify diploma authenticity using a combination of Siamese Neural Network (SNN) for face verification and YOLOv8 for text perforation detection. The SNN model, utilizing a binary cross-entropy loss function, successfully minimized the loss value from 0.169 to 0.0025 by the sixtieth epoch, achieving an accuracy of 0.98. This demonstrates the model's high proficiency in differentiating between anchor and positive/negative images. Additionally, the YOLOv8 model achieved a mean Average Precision (mAP) of 0.93 at a 0.5 IoU threshold, further enhancing the precision of text perforation detection. This integrated system offers a robust and efficient method for ensuring the authenticity of diploma documents, contributing significantly to digital document security and forgery prevention. The research supports the implementation of deep learning in official document verification.