Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2510.26282 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915586809790464 |
|---|---|
| author | Alonso-Fernandez, Fernando Diaz, Kevin Hernandez Buades, Jose M. Raja, Kiran Bigun, Josef |
| author_facet | Alonso-Fernandez, Fernando Diaz, Kevin Hernandez Buades, Jose M. Raja, Kiran Bigun, Josef |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_26282 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Exploring Complementarity and Explainability in CNNs for Periocular Verification Across Acquisition Distances Alonso-Fernandez, Fernando Diaz, Kevin Hernandez Buades, Jose M. Raja, Kiran Bigun, Josef Computer Vision and Pattern Recognition 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. |
| title | Exploring Complementarity and Explainability in CNNs for Periocular Verification Across Acquisition Distances |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.26282 |