...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Sensing and imaging >Transfer Learning for Classification of Optical Satellite Image
【24h】

Transfer Learning for Classification of Optical Satellite Image

机译：转移学习对光学卫星图像的分类

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Deep convolutional neural network (DCNN) has achieved great successrnin the classification of natural images, but it requires numerous labelled datarnfor training. In the absence of a large number of optical satellite images and labelledrndata, how to guarantee the effect of classification of the optical satellite images withrnDCNN? In this case, this paper has discussed how to fine-tune a pre-trained DCNNrnin a layer-wise manner by transfer learning. In our experiment, DCNN is pre-trainedrnwith ImageNet which is a large labelled dataset of natural images, and then opticalrnremote sensing images are used to fine-tune the learnable parameters of pre-trainedrnDCNN. The experimental results show that transfer learning is feasible to deal withrnthe above problem. In the process of transfer training, if the second half of the layersrnare fine-tuned, compared with the fine-tuning of the entire network, the almostrnsame accuracy can be achieved, but the convergence is more rapid. The experimentalrnresults provide a solution for how to achieve the incremental classification performancernin practical applications.
机译：深度卷积神经网络（DCNN）在自然图像分类中取得了巨大的成功，但是它需要大量的标记数据来进行训练。在缺少大量的光学卫星图像和带标签的数据的情况下，如何保证DCNN对光学卫星图像进行分类的效果？在这种情况下，本文讨论了如何通过转移学习以分层的方式微调预训练的DCNNrn。在我们的实验中，对DCNN进行了ImageNet的预训练，ImageNet是一个标记的自然图像的大型数据集，然后使用光学遥感图像微调预训练的DCNN的可学习参数。实验结果表明，转移学习对于解决上述问题是可行的。在传递训练过程中，如果对网络的后半部分进行微调，则与整个网络的微调相比，可以达到几乎相同的精度，但是收敛速度更快。实验结果为实际应用中如何实现增量分类性能提供了解决方案。

著录项

  • 来源
    《Sensing and imaging》 |2018年第1期|7.1-7.13|共13页
  • 作者

    MaoYang Zou; Yong Zhong;

  • 作者单位

    Chengdu Institute of Computer Application, University of Chinese Academy of Sciences, No.9 The 4th Section of South Renmin Road, Chengdu 610041, China Chengdu University of Information Technology, No.24 Block 1, Xuefu Road, Chengdu 610225, China;

    Chengdu Institute of Computer Application, University of Chinese Academy of Sciences, No.9 The 4th Section of South Renmin Road, Chengdu 610041, China;

  • 收录信息
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类
  • 关键词

    Optical remote sensing image; Deep convolutional neural network; Classification; Transfer learning; Fine-tuning;

    机译：光学遥感影像;深度卷积神经网络分类;转移学习;微调;

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号