Title:
MACHINE LEARNING VIA TWO-DIMENSIONAL SYMBOL
Document Type and Number:
Japanese Patent JP2020091831
Kind Code:
A
Abstract:
To provide a method for facilitating machine learning via a two-dimensional symbol.SOLUTION: A method for facilitating machine learning via a 2D symbol includes the steps of: receiving a plurality of features of an object in a first computing system in which a 2D symbol creation application module is installed (502); forming a multilayer 2D symbol from the received features based on a symbol creation rule (504), in which the 2D symbol is a matrix of N×N pixel data (N is a positive integer), the matrix is divided into partial matrices, and each partial matrix has a corresponding one feature; and learning meaning of a composite feature included in the 2D symbol using an image processing technique in a second computing system for classifying the 2D symbol formed in first computing system (506). The symbol creation rule determines an order of importance, size, and a position of the partial matrix within the 2D symbol.SELECTED DRAWING: Figure 5A
Inventors:
YANG LING
BAOHUA SUN
BAOHUA SUN
Application Number:
JP2019132997A
Publication Date:
June 11, 2020
Filing Date:
July 18, 2019
Export Citation:
Assignee:
GYRFALCON TECH INC
International Classes:
G06F40/20; G06N20/00; G06N20/10; G06T7/00; G06V30/40; G06V30/10
Attorney, Agent or Firm:
Hidetada Kato
Masateru Misaki
Masateru Misaki