Presentation 2011-02-17
An Extension of Semantic Texton Forests to Multi-Label Problem Using Label Ranking
Takaaki FUKUTA, Hironobu FUJIYOSHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A method for multi-label semantic segmentation is proposed. The proposed method is based on the use of semantic texton forests, which can compute texton features faster than other method. We adapted the training algorithm for semantic texton forests to multi-label problems by using label ranking, which is a multi-label learning task. Our multi-label semantic segmentation method consists of two steps : extracting features from each pixel of an image by using semantic texton forests that have been adapted to multi-label problems by using label ranking, and classifying each pixel by using multi-label classifiers that have been trained using label powersets, where sets of labels are considered as one category. The proposed method can divide regions into not only single label categories but also multi-label categories. Experimental results for an original dataset including multi-label categories show that our method has about 70% segmentation accuracy.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Semantic Segmentation / Semantic Texton Forest / Multi-Label / Label Ranking
Paper # PRMU2010-210
Date of Issue

Conference Information
Committee PRMU
Conference Date 2011/2/10(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) An Extension of Semantic Texton Forests to Multi-Label Problem Using Label Ranking
Sub Title (in English)
Keyword(1) Semantic Segmentation
Keyword(2) Semantic Texton Forest
Keyword(3) Multi-Label
Keyword(4) Label Ranking
1st Author's Name Takaaki FUKUTA
1st Author's Affiliation Dept. of Computer Science, Chubu University()
2nd Author's Name Hironobu FUJIYOSHI
2nd Author's Affiliation Dept. of Computer Science, Chubu University
Date 2011-02-17
Paper # PRMU2010-210
Volume (vol) vol.110
Number (no) 414
Page pp.pp.-
#Pages 6
Date of Issue