Exploring Complementarity and Explainability in CNNs for Periocular Verification Across Acquisition Distances

TL;DR

This paper investigates how different CNN architectures complement each other in periocular verification across various distances, using explainability tools to understand their focus regions and achieving state-of-the-art results.

Contribution

It introduces a comprehensive analysis of CNN complementarity and explainability in periocular verification, demonstrating improved performance through fusion of diverse models.

Findings

ResNet50 performs best individually.

Fusion of multiple CNNs yields significant accuracy improvements.

Networks focus on different image regions, explaining their complementarity.

Abstract

We study the complementarity of different CNNs for periocular verification at different distances on the UBIPr database. We train three architectures of increasing complexity (SqueezeNet, MobileNetv2, and ResNet50) on a large set of eye crops from VGGFace2. We analyse performance with cosine and chi2 metrics, compare different network initialisations, and apply score-level fusion via logistic regression. In addition, we use LIME heatmaps and Jensen-Shannon divergence to compare attention patterns of the CNNs. While ResNet50 consistently performs best individually, the fusion provides substantial gains, especially when combining all three networks. Heatmaps show that networks usually focus on distinct regions of a given image, which explains their complementarity. Our method significantly outperforms previous works on UBIPr, achieving a new state-of-the-art.