Palmprint Recognition Using Siamese Networks with ResNet-18
Tarek Awad Elbendary, Abeer Twakol Khalil, Mahmoud M. Saafan, Hossam El-Din Mostafa · 2025
Palmprint recognition has emerged as a highly reliable biometric modality due to its unique characteristics, including uniqueness, stability, and high accuracy. Deep learning techniques, particularly Siamese networks, have performed exceptionally in biometric recognition tasks in recent years. This paper presents a comprehensive study on palmprint authentication using Siamese networks with residual network (ResNet-18) as the backbone architecture. The proposed approach leverages the power of deep learning to extract discriminative features from palmprint images. It employs a Siamese network to learn a robust similarity metric between pairs of palmprint images. The ResNet-18 architecture is selected for its optimal balance between model complexity and performance, making it well-suited for palmprint recognition tasks. The proposed method is evaluated on publicly available palmprint datasets, achieving an Equal Error Rate (EER) of 0.0669, demonstrating its effectiveness in achieving high recognition accuracy. Additionally, the paper provides an extensive review of recent literature on palmprint recognition and Siamese networks.