Proposed Method (Training phase)
e.g. SIFT
Labeling Features Proposed-SOINN
…
Attributes Extraction
・・・
Images of Feature-1 F-2 F-Q
Q : Number of Features
Training classes
画像の特徴と属性の関連を学習
Class/Attribute SOINNの数=特徴の数
Matrix
これを利用してクラスから属性へ変換
Images of
Test classes ex) 「りんご」⇒「赤+球」
Dataset
Proposed Method (Test phase)
e.g. SIFT
Labeling Features Proposed-SOINN
…
Attributes Extraction
・・・
Images of Feature-1 F-2 F-Q
Training classes 与えられた特徴から
各々の属性の度合いを算出
Class/Attribute Calculate Attributes using statistical recognition
Matrix
未知クラスも属性のみは知っている
Images of Guess the Class according to these Attributes
Test classes
Q : Number of Features
Dataset 各々の属性からクラスへ変換
ex) 「赤+球」⇒「りんご」
主要参考文献
[1] C. H. Lampert, H. Nickisch, and S. Harmeling, “Learning to detect
unseen object classes by between-class attribute transfer”, CVPR
2009.
[2] A. Kawewong, Sirinart Tangruamsub, Pichai Kankuekool and
Osamu Hasegawa, “Fast Online Incremental Transfer Learning for
Unseen Object Classification Using Self-Organizing Incremental
Neural Networks”, The 2011 International Joint Conference on
Neural Networks (IJCNN).
[3] F. Shen, O. Hasegawa, “An Incremental Network for On-line
Unsupervised Classification and Topology Learning”, Neural
Networks 2006.