Bioinspired CNNs for border completion in occluded images

TL;DR

This paper introduces BorderNet, a CNN architecture inspired by the visual cortex's border completion process, designed to improve robustness to occlusions in images across multiple datasets.

Contribution

The paper presents a novel biologically inspired CNN architecture, BorderNet, specifically tailored for border completion in occluded images, demonstrating improved robustness over existing methods.

Findings

BorderNet outperforms baseline models on MNIST, Fashion-MNIST, and EMNIST datasets.

Performance gains are more significant with increased occlusion severity.

BorderNet effectively handles different types of occlusions, such as stripes and grids.

Abstract

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.