GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond

TL;DR

This paper introduces GCNet, a simplified and efficient global context modeling framework that unifies non-local and squeeze-excitation networks, achieving superior performance on recognition benchmarks.

Contribution

It proposes a query-independent global context block, unifies non-local and squeeze-excitation networks into a general framework, and demonstrates improved efficiency and accuracy in recognition tasks.

Findings

GCNet outperforms NLNet and SENet on benchmarks.

The global context block is lightweight and effective.

Unification of different global context methods into a single framework.

Abstract

The Non-Local Network (NLNet) presents a pioneering approach for capturing long-range dependencies, via aggregating query-specific global context to each query position. However, through a rigorous empirical analysis, we have found that the global contexts modeled by non-local network are almost the same for different query positions within an image. In this paper, we take advantage of this finding to create a simplified network based on a query-independent formulation, which maintains the accuracy of NLNet but with significantly less computation. We further observe that this simplified design shares similar structure with Squeeze-Excitation Network (SENet). Hence we unify them into a three-step general framework for global context modeling. Within the general framework, we design a better instantiation, called the global context (GC) block, which is lightweight and can effectively model the global context. The lightweight property allows us to apply it for multiple layers in a backbone network to construct a global context network (GCNet), which generally outperforms both simplified NLNet and SENet on major benchmarks for various recognition tasks. The code and configurations are released at https://github.com/xvjiarui/GCNet.