Title:
LEARNING DEVICE, LEARNING DEVICE OPERATION PROGRAM, AND LEARNING DEVICE OPERATION METHOD
Document Type and Number:
WIPO Patent Application WO/2020/066233
Kind Code:
A1
Abstract:
This learning device extracts, for each layer, a feature different in the frequency band of spatial frequencies, included in an input image, so as to perform learning for a machine learning model for performing semantic segmentation in which specifications of a plurality of classes in the input image are performed on a pixel-by-pixel unit. In the present invention, a learning data analysis unit analyzes a frequency band included in an annotation image in learning data. A learning method determination unit determines a learning method that uses the learning data, on the basis of the analysis result of the frequency band by the learning data analysis unit. A learning unit learns the machine learning model according to the determined learning method, using the learning data.
Inventors:
WAKUI TAKASHI (JP)
Application Number:
PCT/JP2019/028077
Publication Date:
April 02, 2020
Filing Date:
July 17, 2019
Export Citation:
Assignee:
FUJIFILM CORP (JP)
International Classes:
G06T7/00
Other References:
RONNEBERGER, 0. ET AL.: "U-Net: Convolutional Networks for Biomedical Image Segmentation", ARXIV:1505.04597V1, 18 May 2015 (2015-05-18), pages 1 - 8, XP055562805, Retrieved from the Internet [retrieved on 20190918]
Attorney, Agent or Firm:
NAKASHIMA Junko et al. (JP)
Download PDF: