Biometric Identification using Periocular Images with ViT-DeepSort and YOLOv7-GAN

Advanced biometric identification techniques are becoming increasingly necessary in response to rising security requirements. The limits of conventional facial recognition systems have been further brought to light by the COVID-19 outbreak and its emphasis on mask-wearing. utilizing periocular pictures, this study proposes an integrated technique to improve gender categorization and biometric identification utilizing Vision Transformers (ViTs), YOLOv7, DeepSort, and Generative Adversarial Networks (GANs). Our findings indicate that the combined application of these technologies results in a system that is both robust and accurate, particularly in challenging environments. The proposed methodology demonstrates a notable improvement in identification accuracy, with a precision rate exceeding $\mathbf{9 5 \%}$. With this research, secure identification technologies will advance significantly, and non-intrusive, contactless biometric devices will continue to be developed. The findings suggest that the application of these cutting-edge techniques has the potential to completely transform security procedures in industries including public safety, healthcare, and finance.