Bioinspired CNNs for border completion in occluded images
This addresses image occlusion robustness for computer vision applications, but appears incremental as it builds on existing CNN methods with bioinspired filters.
The paper tackled the problem of border completion in occluded images by designing a bioinspired CNN architecture, BorderNet, which showed improved performance on occluded datasets like MNIST, Fashion-MNIST, and EMNIST under stripe and grid occlusions, with gains varying based on occlusion severity and dataset.
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.