Related Experiment Video
Updated: Jan 13, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
IRIS-QResNet: A Quantum-Inspired Deep Model for Efficient Iris Biometric Identification and Authentication
Neama Abdulaziz Dahan1, Emad Sami Jaha1
1Department of Computer Science, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah 21589, Saudi Arabia.
Sensors (Basel, Switzerland)
|January 10, 2026
Summary
This study introduces IRIS-QResNet, a quantum-inspired deep learning model that enhances iris recognition accuracy, especially with limited data. The novel quanvolutional layer significantly improves feature extraction and recognition performance over traditional ResNet models.
Area of Science:
- Computer Science
- Artificial Intelligence
- Quantum Computing
Background:
- Iris recognition is a reliable biometric technique facing challenges with deep learning models, particularly when training data is limited.
- Subtle, high-dimensional patterns crucial for accurate iris recognition are often missed by conventional models with insufficient data.
Purpose of the Study:
- To enhance feature extraction and recognition accuracy in iris recognition systems using deep learning.
- To address the limitations of current models in handling limited training data for iris recognition.
Main Methods:
- Introduction of IRIS-QResNet, a customized ResNet-18 architecture incorporating a quanvolutional layer.
- The quanvolutional layer simulates quantum effects (entanglement, superposition) and uses sinusoidal feature encoding for improved multilayer representations.
- 14 experiments were conducted across four datasets (CASIA-Thousands, IITD, MMU, UBIris) comparing IRIS-QResNet against an IResNet baseline.
Main Results:
- IRIS-QResNet consistently outperformed the IResNet baseline in accuracy and reduced loss across multiple datasets and recognition tasks.
- Accuracy improvements for IRIS-QResNet ranged from 0.1870% to 16.67%, with loss reductions varying from 0.0360 to 1.0280.
- The model demonstrated robust generalization capabilities even without data augmentation, highlighting the effectiveness of quantum-inspired modifications.
Conclusions:
- Quantum-inspired modifications, specifically the quanvolutional layer, offer a practical and scalable method to boost the discriminative power of residual networks.
- The proposed IRIS-QResNet effectively bridges classical deep learning with emerging quantum machine learning paradigms for enhanced biometric security.
- This approach shows significant promise for improving the performance of iris recognition systems in real-world, data-constrained scenarios.
