Instance-Level Meta Normalization

Jia, Songhao; Chen, Ding-Jie; Chen, Hwann-Tzong

doi:10.1109/CVPR.2019.00500

Instance-Level Meta Normalization

Songhao Jia, Ding-Jie Chen, Hwann-Tzong Chen

CVPR 2019

doi:10.1109/CVPR.2019.00500 /cvpr/2019/jia2019cvpr-instancelevel/

Abstract

This paper presents a normalization mechanism called Instance-Level Meta Normalization (ILM Norm) to address a learning-to-normalize problem. ILM Norm learns to predict the normalization parameters via both the feature feed-forward and the gradient back-propagation paths. ILM Norm provides a meta normalization mechanism and has several good properties. It can be easily plugged into existing instance-level normalization schemes such as Instance Normalization, Layer Normalization, or Group Normalization. ILM Norm normalizes each instance individually and therefore maintains high performance even when small mini-batch is used. The experimental results show that ILM Norm well adapts to different network architectures and tasks, and it consistently improves the performance of the original models.

PDF CVPR Semantic Scholar

Cite

Text

Jia et al. "Instance-Level Meta Normalization." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019. doi:10.1109/CVPR.2019.00500

Markdown

[Jia et al. "Instance-Level Meta Normalization." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019.](https://mlanthology.org/cvpr/2019/jia2019cvpr-instancelevel/) doi:10.1109/CVPR.2019.00500

BibTeX

@inproceedings{jia2019cvpr-instancelevel,
  title     = {{Instance-Level Meta Normalization}},
  author    = {Jia, Songhao and Chen, Ding-Jie and Chen, Hwann-Tzong},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year      = {2019},
  doi       = {10.1109/CVPR.2019.00500},
  url       = {https://mlanthology.org/cvpr/2019/jia2019cvpr-instancelevel/}
}