Publication detail

Comparing Normalization Methods for Limited Batch Size Segmentation Neural Networks

KOLAŘÍK, M. BURGET, R. ŘÍHA, K.

Original Title

Comparing Normalization Methods for Limited Batch Size Segmentation Neural Networks

Type

conference paper

Language

English

Original Abstract

The widespread use of Batch Normalization has enabled training deeper neural networks with more stable and faster results. However, the Batch Normalization works best using large batch size during training and as the state-of-theart segmentation convolutional neural network architectures are very memory demanding, large batch size is often impossible to achieve on current hardware. We evaluate the alternative normalization methods proposed to solve this issue on a problem of binary spine segmentation from 3D CT scan. Our results show the effectiveness of Instance Normalization in the limited batch size neural network training environment. Out of all the compared methods the Instance Normalization achieved the highest result with Dice coefficient = 0.96 which is comparable to our previous results achieved by deeper network with longer training time. We also show that the Instance Normalization implementation used in this experiment is computational timeefficient when compared to the network without any normalization method.

Keywords

Batch Normalization; Comparison; Group Normalization; Instance Normalization; Segmentation

Authors

KOLAŘÍK, M.; BURGET, R.; ŘÍHA, K.

Released

9. 7. 2020

ISBN

978-1-7281-6376-5

Book

2020 43rd International Conference on Telecommunications and Signal Processing (TSP)

Pages from

677

Pages to

680

Pages count

4

URL

BibTex

@inproceedings{BUT164792,
  author="Martin {Kolařík} and Radim {Burget} and Kamil {Říha}",
  title="Comparing Normalization Methods for Limited Batch Size Segmentation Neural Networks",
  booktitle="2020 43rd International Conference on Telecommunications and Signal Processing (TSP)",
  year="2020",
  pages="677--680",
  doi="10.1109/TSP49548.2020.9163397",
  isbn="978-1-7281-6376-5",
  url="https://ieeexplore.ieee.org/document/9163397"
}