Presentation 2021-12-16
Verification of Cyclical Annealing for Object-Oriented Representation Learning
Atsushi Kobayashi, Hideki Tsunashima, Takehiko Ohkawa, Hiroaki Aizawa, Qiu Yue, Hirokatsu Kataoka, Shigeo Morishima,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Object-oriented Representation Learning is a method for obtaining images for each object and background part from an image. This method is expected to improve the performance of downstream tasks that require a structural understanding of the scene. However, the existing methods for object-aware representation learning have a problem that the decomposition of background parts is sometimes not performed correctly. In a previous study, the cause of this problem was shown to be the loss of KL-divergence between the posterior and prior distributions of the latent variable, and a solution was proposed that periodically increases or decreases the loss for the latent variable during learning. In this study, we quantitatively verify the effectiveness of this learning method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Object-oriented Representation Learning / Variational Auto-Encoder / Unsupervised Learning
Paper # PRMU2021-39
Date of Issue 2021-12-09 (PRMU)

Conference Information
Committee PRMU
Conference Date 2021/12/16(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Online
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Seiichi Uchida(Kyushu Univ.)
Vice Chair Masakazu Iwamura(Osaka Pref. Univ.) / Mitsuru Anpai(Denso IT Lab.)
Secretary Masakazu Iwamura(NTT) / Mitsuru Anpai(Tottori Univ.)
Assistant Kouta Yamaguchi(CyberAgent) / Yusuke Matsui(Univ. of Tokyo)

Paper Information
Registration To Technical Committee on Pattern Recognition and Media Understanding
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Verification of Cyclical Annealing for Object-Oriented Representation Learning
Sub Title (in English)
Keyword(1) Object-oriented Representation Learning
Keyword(2) Variational Auto-Encoder
Keyword(3) Unsupervised Learning
1st Author's Name Atsushi Kobayashi
1st Author's Affiliation Waseda University(Waseda Univ.)
2nd Author's Name Hideki Tsunashima
2nd Author's Affiliation Waseda University/National Institute of Advanced Industrial Science and Technology(Waseda Univ./AIST)
3rd Author's Name Takehiko Ohkawa
3rd Author's Affiliation The University of Tokyo(The Univ. of Tokyo)
4th Author's Name Hiroaki Aizawa
4th Author's Affiliation Hiroshima University(Hiroshima Univ.)
5th Author's Name Qiu Yue
5th Author's Affiliation National Institute of Advanced Industrial Science and Technology(AIST)
6th Author's Name Hirokatsu Kataoka
6th Author's Affiliation National Institute of Advanced Industrial Science and Technology(AIST)
7th Author's Name Shigeo Morishima
7th Author's Affiliation Waseda University(Waseda Univ.)
Date 2021-12-16
Paper # PRMU2021-39
Volume (vol) vol.121
Number (no) PRMU-304
Page pp.pp.83-87(PRMU),
#Pages 5
Date of Issue 2021-12-09 (PRMU)