Siamese Iris on Siamese Neural Network Architecture for Iris Matching and Verification
K. Sivasankari, D. Kerana Hanirex · 2025
This chapter explores iris identification challenges, including occlusions, distortions, and tex-tural variations. A novel architecture for Siamese neural networks called SiameseIris was de-veloped to identify and verify the iris biometrics. There is great anticipation surrounding the potential revolution in iris matching and verification. The execution of SiameseIris on the CASIA-IrisV4 dataset yields an exceptional accuracy rate of 99.06%. The model utilizes a Sia-mese network architecture consisting of two subnetworks that share weights to identify similari-ties between pairs of iris images. Extensive experimental evaluation has confirmed the impres-sive performance of SiameseIris in iris recognition tasks, surpassing existing methods. The model's reliability and durability are supported by its impressive FAR and FRR scores of 0.49 and 0.65 in accurately matching iris patterns. Biometric authentication systems, such as Sia-meseIris, have enhanced access control, identity verification, and forensic investigation security and reliability.