Unsupervised Techniques for the Propagation of Label Information in Image Datasets

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500135
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Media informatics/Database
• Research Institution: National Institute of Informatics
• Principal Investigator: HOULE Michael E. 国立情報学研究所, 大学共同利用機関等の部局等, 客員教授 (90399270)
• Co-Investigator(Kenkyū-buntansha): SATOH Shin'ichi 国立情報学研究所, コンテンツ科学研究系, 教授 (90249938)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: ラベル伝播 / 分類アルゴリズム / 類似検索 / 画像データ

Outline of Final Research Achievements

Despite much effort by researchers, the problem of automatic recognition of face images remains very difficult. When processing face image queries, the most effective of the current face recognition algorithms determine as the query result a set of images that most closely resemble several of a fixed set of labeled examples. Typically, image databases contain multiple images that may be similar to a given query. The goal of this project is to increase the number of database images with label information, by sharing face image annotations with unlabeled images that are similar to them.

Free Research Field

データマイニング