Skip to content

The official implementation of paper "Instance Enhancement Batch Normalization: an Adaptive Regulator of Batch Noise".

License

Notifications You must be signed in to change notification settings

gbup-group/IEBN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Jun 16, 2023
ed6d563 · Jun 16, 2023

History

35 Commits
Aug 13, 2019
Aug 13, 2019
Aug 13, 2019
Oct 23, 2019
Jun 16, 2023
Aug 13, 2019

Repository files navigation

Instance Enhancement Batch Normalization: an Adaptive Regulator of Batch Noise

GitHub GitHub

By Senwei Liang*, Zhongzhan Huang* (* contribute equally), Mingfu Liang and Haizhao Yang.

This repository is the implementation of "Instance Enhancement Batch Normalization: an Adaptive Regulator of Batch Noise" [paper] on CIFAR-100 dataset. Our paper has been accepted for presentation at the Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI-20). You can also check with the AAAI proceeding version.

Introduction

Instance Enhancement Batch Normalization (IEBN) is an attention-based version of BN which recalibrates channel information of BN by a simple linear transformation.

Requirement

Usage

python cifar.py -a iebn_resnet --dataset cifar100 --block-name bottleneck --depth 164 --epochs 164 --schedule 81 122 --gamma 0.1 --wd 1e-4 --checkpoint checkpoints/cifar100/resnet-164-iebn

Results

original IEBN
ResNet164 74.29 77.09

Notes:

  • Training on 2 GPUs

Citing IEBN

@inproceedings{liang2020instance,
  title={Instance Enhancement Batch Normalization: An Adaptive Regulator of Batch Noise.},
  author={Liang, Senwei and Huang, Zhongzhan and Liang, Mingfu and Yang, Haizhao},
  booktitle={AAAI},
  pages={4819--4827},
  year={2020}
}

Acknowledgments

Many thanks to bearpaw for his simple and clean Pytorch framework for image classification task.

About

The official implementation of paper "Instance Enhancement Batch Normalization: an Adaptive Regulator of Batch Noise".

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages