Enhancing CNNs robustness to occlusions with bioinspired filters for border completion
Main:6 Pages
6 Figures
Bibliography:2 Pages
Abstract
We exploit the mathematical modeling of the visual cortex mechanism for border completion to define custom filters for CNNs. We see a consistent improvement in performance, particularly in accuracy, when our modified LeNet 5 is tested with occluded MNIST images.
View on arXiv@article{coutinho2025_2504.17619,
title={ Enhancing CNNs robustness to occlusions with bioinspired filters for border completion },
author={ Catarina P. Coutinho and Aneeqa Merhab and Janko Petkovic and Ferdinando Zanchetta and Rita Fioresi },
journal={arXiv preprint arXiv:2504.17619},
year={ 2025 }
}Comments on this paper