Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ResNeXt:Aggregated Residual Transformations for Deep Neural Networks #28

Open
chullhwan-song opened this issue Jul 26, 2018 · 1 comment
Labels

Comments

@chullhwan-song
Copy link
Owner

https://github.com/facebookresearch/ResNeXt

@chullhwan-song
Copy link
Owner Author

  • Author : Kaiming He외 4명

  • 특징

    • cardinality(관계수, 1:n 관계?)라 불리우는 형태를 가진다.
      • cardinality = the size of the set of transformations
      • 밑의 그림에서 보면 기본적 구조는 resnet과 유사하나 resnet의 각각의 Layer를 조각(channel)낸 형태를 가진다. 이를 나중에 cat한다.
        • 이때의 cardinality = 32
          image
  • 기본적으로 resnet과 유사(또는 비교), 그리고 성능향상되고, 반면에 거의 같은 complexity(paraemeter, FLOPS)를 가진다.

    • FLOPS : 플롭스(FLOPS, FLoating point OPerations per Second)는 컴퓨터의 성능을 수치로 나타낼 때 주로 사용되는 단위 : link
  • 성능향상
    image

    • 1%의 향상을 이루어냄
      • ImageNet-1K (top-1 error)
        • ResNet-50 vs ResNeXt-50 = 23.9 vs 22.2
        • ResNet-101 vs ResNeXt-101= 22.0 vs 21.2

@chullhwan-song chullhwan-song changed the title Aggregated Residual Transformations for Deep Neural Networks ResNeXt:Aggregated Residual Transformations for Deep Neural Networks Aug 23, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant