Bioinspired CNNs for border completion in occluded images

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Coutinho, Catarina P., Merhab, Aneeqa, Petkovic, Janko, Zanchetta, Ferdinando, Fioresi, Rita
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914386308759552
author Coutinho, Catarina P.
Merhab, Aneeqa
Petkovic, Janko
Zanchetta, Ferdinando
Fioresi, Rita
author_facet Coutinho, Catarina P.
Merhab, Aneeqa
Petkovic, Janko
Zanchetta, Ferdinando
Fioresi, Rita
contents We exploit the mathematical modeling of the border completion problem in the visual cortex to design convolutional neural network (CNN) filters that enhance robustness to image occlusions. We evaluate our CNN architecture, BorderNet, on three occluded datasets (MNIST, Fashion-MNIST, and EMNIST) under two types of occlusions: stripes and grids. In all cases, BorderNet demonstrates improved performance, with gains varying depending on the severity of the occlusions and the dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2603_10694
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Bioinspired CNNs for border completion in occluded images
Coutinho, Catarina P.
Merhab, Aneeqa
Petkovic, Janko
Zanchetta, Ferdinando
Fioresi, Rita
Computer Vision and Pattern Recognition
We exploit the mathematical modeling of the border completion problem in the visual cortex to design convolutional neural network (CNN) filters that enhance robustness to image occlusions. We evaluate our CNN architecture, BorderNet, on three occluded datasets (MNIST, Fashion-MNIST, and EMNIST) under two types of occlusions: stripes and grids. In all cases, BorderNet demonstrates improved performance, with gains varying depending on the severity of the occlusions and the dataset.
title Bioinspired CNNs for border completion in occluded images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.10694