Gather-Excite: Exploiting Feature Context in Convolutional Neural Networks

TL;DR

Gather-Excite introduces lightweight operators for CNNs that efficiently capture long-range feature interactions, significantly improving performance without increasing model complexity.

Contribution

The paper proposes a novel, simple gather-excite operator pair for CNNs that enhances context exploitation and improves accuracy with minimal additional parameters.

Findings

Gather-Excite improves CNN performance on multiple datasets.

ResNet-50 with Gather-Excite outperforms deeper ResNet-101.

Parametric Gather-Excite yields further gains.

Abstract

While the use of bottom-up local operators in convolutional neural networks (CNNs) matches well some of the statistics of natural images, it may also prevent such models from capturing contextual long-range feature interactions. In this work, we propose a simple, lightweight approach for better context exploitation in CNNs. We do so by introducing a pair of operators: gather, which efficiently aggregates feature responses from a large spatial extent, and excite, which redistributes the pooled information to local features. The operators are cheap, both in terms of number of added parameters and computational complexity, and can be integrated directly in existing architectures to improve their performance. Experiments on several datasets show that gather-excite can bring benefits comparable to increasing the depth of a CNN at a fraction of the cost. For example, we find ResNet-50 with gather-excite operators is able to outperform its 101-layer counterpart on ImageNet with no additional learnable parameters. We also propose a parametric gather-excite operator pair which yields further performance gains, relate it to the recently-introduced Squeeze-and-Excitation Networks, and analyse the effects of these changes to the CNN feature activation statistics.