GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond
This work improves efficiency and performance for computer vision models by addressing redundant computations in capturing long-range dependencies, though it is incremental as it builds on existing NLNet and SENet approaches.
The authors identified that non-local networks produce nearly identical global contexts across different query positions, leading them to develop a simplified, query-independent formulation that matches NLNet's accuracy with less computation. They unified this with Squeeze-Excitation Networks into a general framework and introduced the GC block, which outperforms both simplified NLNet and SENet on major benchmarks for recognition tasks.
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.