RT Journal Article
T1 Enhanced robustness of convolutional networks with a push-pull inhibition layer
A1 Strisciuglio, Nicola
A1 Lopez-Antequera, Manuel
A1 Petkov, Nicolai
K1 Convolutional neural networks
K1 Image corruption
K1 Network robustness
K1 Neuron response inhibition
K1 push-pull layer
K1 Orientation selectivity
K1 Receptive-fields
K1 Simple cell
K1 Suppression
K1 Broad
K1 Model
AB Convolutional neural networks (CNNs) lack robustness to test image corruptions that are not seen during training. In this paper, we propose a new layer for CNNs that increases their robustness to several types of corruptions of the input images. We call it a 'push-pull' layer and compute its response as the combination of two half-wave rectified convolutions, with kernels of different size and opposite polarity. Its implementation is based on a biologically motivated model of certain neurons in the visual system that exhibit response suppression, known as push-pull inhibition. We validate our method by replacing the first convolutional layer of the LeNet, ResNet and DenseNet architectures with our push-pull layer. We train the networks on original training images from the MNIST and CIFAR data sets and test them on images with several corruptions, of different types and severities, that are unseen by the training process. We experiment with various configurations of the ResNet and DenseNet models on a benchmark test set with typical image corruptions constructed on the CIFAR test images. We demonstrate that our push-pull layer contributes to a considerable improvement in robustness of classification of corrupted images, while maintaining state-of-the-art performance on the original image classification task. We released the code and trained models at the url http://github.com/nicstrisc/Push-Pull-CNN-layer.
PB Springer london ltd
SN 0941-0643
YR 2020
FD 2020-01-23
LK http://hdl.handle.net/10668/18637
UL http://hdl.handle.net/10668/18637
LA en
NO Strisciuglio N, Lopez-Antequera M, Petkov N. Enhanced robustness of convolutional networks with a push–pull inhibition layer. Neural Comput Appl. 2020;32:17957–17971
NO This study was partially funded by the European Commission H2020 program under project TrimBot2020 (Grant Number 688007).
DS RISalud
RD Apr 6, 2025