Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
IDENTIFICATION DICTIONARY LEARNING SYSTEM, IDENTIFICATION DICTIONARY LEARNING METHOD, AND RECORDING MEDIUM
Document Type and Number:
WIPO Patent Application WO/2015/146113
Kind Code:
A1
Abstract:
Provided is an identification dictionary learning system and the like that create a dictionary which is used in identification and with which a high identification rate can be obtained, in order to detect a prescribed object for which there is a large apparent amount of variation in an image. This identification dictionary learning system (200) is equipped with: a setting unit (211), which receives priorities for a classification comprising a first class and a second class that includes subclasses, and which sets those priorities; an identifier learning unit (213) which, for each combination of subclasses, learns multiple identifiers on the basis of the priorities; a selection unit (215), which selects the combination of subclasses for which a prescribed optimization index for an identifier is optimal; and a weight updating unit (216), which updates a weight on the basis of an identification precision by means of the identifier learned with the selected combination.

Inventors:
HOSOI TOSHINORI (JP)
Application Number:
PCT/JP2015/001594
Publication Date:
October 01, 2015
Filing Date:
March 20, 2015
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NEC CORP (JP)
International Classes:
G06T7/00; G06N20/00; G06V10/772
Domestic Patent References:
WO2013165336A12013-11-07
Foreign References:
JP2011198181A2011-10-06
JP2013033331A2013-02-14
JP2008250908A2008-10-16
JP2008217589A2008-09-18
JP2011027498A2011-02-10
US20050209785A12005-09-22
Other References:
ASAKO KANEZAKI ET AL.: "Daikibo Gazo Dataset o Mochiita Multi Class Buttai Kenshutsuki no Doji Gakushu", IPSJ SIG NOTES COMPUTER VISION AND IMAGE MEDIA (CVIM, vol. 2012 -CV, no. 12, 15 October 2012 (2012-10-15), pages 1 - 8
Attorney, Agent or Firm:
SHIMOSAKA, NAOKI (JP)
Naoki Shimosaka (JP)
Download PDF: