arXiv 2510.262822025-10-30cs.CV

探索CNN在跨采集距离眼周验证中的互补性与可解释性

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

School of Information Technology Halmstad University, Sweden（信息科技学院哈姆斯塔德大学，瑞典）；AI Group University of Balearic Islands, Spain（人工智能小组巴利阿里群岛大学，西班牙）；Faculty of Information Technology（信息科技学院）

Fernando Alonso-Fernandez, Kevin Hernandez Diaz, Jose M. Buades, Kiran Raja, Josef Bigun

备注：: Accepted at BIOSIG 2025 conference

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文研究不同CNN在跨采集距离眼周验证中的互补性，通过训练SqueezeNet、MobileNetv2和ResNet50并结合分数级融合，利用LIME热图解释其注意力差异，在UBIPr上达到新SOTA。

摘要：

我们在UBIPr数据库上研究不同CNN在跨采集距离眼周验证中的互补性。我们在来自VGGFace2的大量眼部裁剪图像上训练了三种复杂度递增的架构（SqueezeNet、MobileNetv2和ResNet50）。我们使用余弦和chi2度量分析性能，比较不同的网络初始化，并通过逻辑回归应用分数级融合。此外，我们使用LIME热图和Jensen-Shannon散度来比较CNN的注意力模式。虽然ResNet50单独表现始终最佳，但融合带来了显著提升，尤其是在结合所有三个网络时。热图显示网络通常关注给定图像的不同区域，这解释了它们的互补性。我们的方法显著优于UBIPr上的先前工作，达到了新的最先进水平。

英文摘要：

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.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。