Presentation | 2005-10-28 Gender Recognition using multiple classifiers trained with sets of clustered features Takahiko KUWABARA, Hitoshi IKEDA, Noriji KATO, Hirotsugu KASHIMURA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Recognition of person's attributes (sex, age, and race, etc.) from the person's face image using machine learning technique has been researched. Person attribute recognition rate in real environment decreases greatly compared with that in controlled environment, because difference of face images caused from changes of lighting condition and face direction is usually larger than that caused from the attribute difference. In this research, we try to improve recognition rate under real environment using multiple classifiers with high recognition performance trained under specific environment. Feature vectors for training are extracted from face images taken in various environments, and divided to multiple groups by using the clustering technique. The feature vectors in each group are used to train individual classifiers. Moreover, performance degradation caused from cluster boundary is prevented by sharing training samples near the cluster boundary with related classifiers. Our method achieved gender recognition rate of 87.3% in real environment with 4-directional surface feature and a support vector machine. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Face Image / Gender recognition / Clustering |
Paper # | PRMU2005-93 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 2005/10/21(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) | Gender Recognition using multiple classifiers trained with sets of clustered features |
Sub Title (in English) | |
Keyword(1) | Face Image |
Keyword(2) | Gender recognition |
Keyword(3) | Clustering |
1st Author's Name | Takahiko KUWABARA |
1st Author's Affiliation | Corporate Research Lab., Fuji Xerox Co., Ltd.() |
2nd Author's Name | Hitoshi IKEDA |
2nd Author's Affiliation | Corporate Research Lab., Fuji Xerox Co., Ltd. |
3rd Author's Name | Noriji KATO |
3rd Author's Affiliation | Corporate Research Lab., Fuji Xerox Co., Ltd. |
4th Author's Name | Hirotsugu KASHIMURA |
4th Author's Affiliation | Corporate Research Lab., Fuji Xerox Co., Ltd. |
Date | 2005-10-28 |
Paper # | PRMU2005-93 |
Volume (vol) | vol.105 |
Number (no) | 375 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |