Presentation 2012-10-04
Efficient Labeling Method Using Clustering For Image Recognition
Ryo SHIRASU, Jien KATO, Kenji MASE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recent years, with the rapid popularization of digital cameras, the amount of photo data that individuals can keep has grown to a huge amount. In order to organize them efficiently, it is necessary to develop methods and technologies that can recognize their contents. In this work, we focus on reducing the manual labeling effort that is necessary for such image recognition task. The key idea is that features that are characteristic for the "common classes" are also discriminative for classifying "specific class", and can help to find the photo which is "worth for labeling". Our proposed method consists of two steps: 1) select the features that are effective in classifying "common classes" through SVM training; and 2) use the selected features to find photos that are "worth for labeling" in the photo album through clustering. We tested the proposed method in image recognition task on a large photo album which consists of 21424 photos taken from a kindergarten. The result confirmed that the proposed method outperform the baseline methods under different experiment setting.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Image Recognition / Clustering / Labeling / Feature Selection / Active Learning
Paper # PRMU2012-51
Date of Issue

Conference Information
Committee PRMU
Conference Date 2012/9/27(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Efficient Labeling Method Using Clustering For Image Recognition
Sub Title (in English)
Keyword(1) Image Recognition
Keyword(2) Clustering
Keyword(3) Labeling
Keyword(4) Feature Selection
Keyword(5) Active Learning
1st Author's Name Ryo SHIRASU
1st Author's Affiliation Graduate School of Information Science Nagoya University()
2nd Author's Name Jien KATO
2nd Author's Affiliation Graduate School of Information Science Nagoya University
3rd Author's Name Kenji MASE
3rd Author's Affiliation Graduate School of Information Science Nagoya University
Date 2012-10-04
Paper # PRMU2012-51
Volume (vol) vol.112
Number (no) 225
Page pp.pp.-
#Pages 6
Date of Issue