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

**Authors:** Yue Cao, Jiarui Xu, Stephen Lin, Fangyun Wei, Han Hu  
**arXiv:** [1904.11492](https://arxiv.org/abs/1904.11492)  
**Date:** 2019-04-25  
**Summary page:** https://www.summarizepaper.com/en/paper/1904.11492v1/

## Key points

- Authors introduce a novel approach to global context modeling in deep neural networks
- Three-step general framework for efficient global context modeling by unifying NLNet and SENet approaches
- Introduction of a new lightweight instantiation called the GC block effective in capturing long-range dependencies
- Extensive experimentation on various recognition tasks showing GCNet outperforms simplified NLNet and SENet consistently
- Availability of code and configurations for further exploration and implementation at https://github.com/xvjiarui/GCNet

## Summary

In their paper titled "GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond," authors Yue Cao, Jiarui Xu, Stephen Lin, Fangyun Wei, and Han Hu introduce a novel approach to global context modeling in deep neural networks. The authors present a three-step general framework for efficient global context modeling by unifying the Non-Local Network (NLNet) and Squeeze-Excitation Network (SENet) approaches. They also introduce a new lightweight instantiation called the global context (GC) block which is effective in capturing long-range dependencies. Through extensive experimentation on various recognition tasks using major benchmarks, their proposed Global Context Network (GCNet) consistently outperforms both simplified NLNet and SENet. This paper offers valuable insights into improving the efficiency of global context modeling in deep learning architectures. The authors have made their code and configurations available at https://github.com/xvjiarui/GCNet for further exploration and implementation by the research community.