Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
LEARNING DEVICE, SYSTEM FOR GENERATING CAPTURED IMAGE CLASSIFICATION DEVICE, DEVICE FOR GENERATING CAPTURED IMAGE CLASSIFICATION DEVICE, LEARNING METHOD, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2019/146057
Kind Code:
A1
Abstract:
Provided are a learning device, a learning method, and a program capable of improving the accuracy of classifying an image belonging to a target domain. A classification learning unit (74) trains a feature quantity extraction unit (72) and a classification unit (32) on the basis of the result of a comparison between the output from the classification unit (32) as obtained when feature quantity data is input to the classification unit (32) and training data for a plurality of classes associated with a source domain learning image. A dividing unit (50) divides feature quantity data that is output from the feature quantity extraction unit (72) when an image is input to the feature quantity extraction unit (72) into a plurality of pieces of partial feature quantity data that are associated with the image and that each include one or a plurality of class feature maps. A domain identification learning unit (76) trains the feature quantity extraction unit (72) on the basis of the result of a comparison between the output from a domain identification unit (56) as obtained when partial feature quantity data corresponding to an image is input to the domain identification unit (56) and data that indicates whether the image belongs to the source domain or target domain.

Inventors:
ONO DAICHI (JP)
Application Number:
PCT/JP2018/002409
Publication Date:
August 01, 2019
Filing Date:
January 26, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
SONY INTERACTIVE ENTERTAINMENT INC (JP)
International Classes:
G06T7/00; G06N99/00; G06V10/764
Foreign References:
JP2016139176A2016-08-04
JP2014137756A2014-07-28
Other References:
NAKAYAMA, HIDEKI: "Image feature extraction and transfer learning using deep convolutional neural networks", IEICE TECHNICAL REPORT, vol. 115, no. 146, 9 July 2015 (2015-07-09), pages 55 - 59
YSASAKI6023, CHAINER FINE-TUNING, 31 December 2016 (2016-12-31), Retrieved from the Internet [retrieved on 20180418]
Attorney, Agent or Firm:
HARUKA PATENT & TRADEMARK ATTORNEYS (JP)
Download PDF: