The Techniques for Improving Classification Accuracy with Deep Learning

DOI

Bibliographic Information

Other Title
  • 深層学習を用いた画像識別タスクの精度向上テクニック

Search this article

Description

Since AlexNet using deep learning archived the greatly improved classification accuracy, the speed of development is remarkable, as the new methods is published in arXiv every day. However, deep learning require too much computational costs in training, tuning much hyper-parameters and data augmentations for better classification accuracy. In this paper, we aim sharing the knowledges for improving accuracy by the survey about data augmentations, learning rate scheduling and ensembles methods as techniques and verify the effects on it. Finally we conduct the integrated experiments using the techniques that improve especially and describe the areas of future research.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top