[Paper Review] GCNet(2019), Non-local Networks Meet Squeeze-Excitation Networks and Beyond

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

Yue Cao, Jiarui Xu, Stephen Lin, Fangyum Wei, Han Hu, arXiv 2019

 

PDF, Video By SeonghoonYu July 27th, 2021

 

Summary

 

 This paper observes that the global contexts modeled by non-local network are almost the same for different query positions within an image. They calculate the global context abount only one query because calculating all the global contexts about every query positions is very expensive. And then they apply SENet architecture. Finally they chieves SOTA perpormance by applying GCNet which is small complexity to all layers

 

 

 

Experiment

 

 

What I like about the paper

• use lower complexity block than previous SOTA model Non-label networks

---

my github about what i read

Seonghoon-Yu/Paper_Review_and_Implementation_in_PyTorch