首页> 外国专利> LEARNING METHOD AND LEARNING DEVICE FOR IMAGE SEGMENTATION, AND IMAGE SEGMENTATION METHOD AND IMAGE SEGMENTATION DEVICE USING THE SAME

LEARNING METHOD AND LEARNING DEVICE FOR IMAGE SEGMENTATION, AND IMAGE SEGMENTATION METHOD AND IMAGE SEGMENTATION DEVICE USING THE SAME

机译:用于图像分割的学习方法和学习装置,以及使用该方法的图像分割方法和图像分割装置

摘要

To provide a learning method and a testing method for image segmentation, and a learning device and a testing device using the methods.SOLUTION: A learning method belonging to an image recognition technology field using deep learning includes: performing a process of sequentially acquiring (2-K)-th, (2-(K-1))-th, ..., (2-1)-th feature maps via a decoding layer; acquiring feature maps by using feature maps from previous filters of the decoding layer and (3-1)-th, (3-2)-th, ..., (3-H)-th feature maps when acquiring the feature maps from H filters of the decoding layer corresponding to (3-1)-th to (3-H)-th filters; and performing a process of acquiring feature maps by using feature maps from the previous filters of the decoding layer when acquiring the feature maps from (K-H) filters of the decoding layer not corresponding to the (3-1)-th to (3-H)-th filters.SELECTED DRAWING: Figure 2A
机译:提供一种用于图像分割的学习方法和测试方法,以及使用该方法的学习设备和测试设备。解决方案:一种利用深度学习的图像识别技术领域的学习方法,包括:执行顺序获取(2经由解码层的第-K)个,(2-(K-1)),...,(2-1)个特征图;通过从解码层的先前过滤器中获取特征图来获取特征图,并在从中获取特征图时使用第(3-1),(3-2),...,(3-H)个特征图解码层的H个滤波器对应于第(3-1)至第(3-H)滤波器;当从解码层的与第(3-1)到(3-H)不对应的(KH)个滤波器中获取特征图时,通过使用来自解码层的先前滤波器的特征图来获取特征图的过程)过滤器。选定的图:图2A

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

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