Presentation 2020-03-11
Pre-training for Action Classification Task Using Video Frame Prediction Task
Hidemoto Nakada, Hideki Asoh,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Continuous Video frames have strongly correlated with each other and thus include rich information that could be leveraged as training data without having explicit labels. We report a trial of transfer learning, where we employ the future frame prediction task for pre-training and apply it to the video behavior classification task. We confirmed that we could gain a certain level of improvement with the proposed pre-training. The improvement was smaller than the pre-training using the other video classification task, however.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) pre-training / action classification / video frame prediction / self supervised learning
Paper # IBISML2019-45
Date of Issue 2020-03-03 (IBISML)

Conference Information
Committee IBISML
Conference Date 2020/3/10(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Kyoto University
Topics (in Japanese) (See Japanese page)
Topics (in English) Machine learning, etc.
Chair Hisashi Kashima(Kyoto Univ.)
Vice Chair Masashi Sugiyama(Univ. of Tokyo) / Koji Tsuda(Univ. of Tokyo)
Secretary Masashi Sugiyama(Nagoya Inst. of Tech.) / Koji Tsuda(AIST)
Assistant Tomoharu Iwata(NTT) / Shigeyuki Oba(Kyoto Univ.)

Paper Information
Registration To Technical Committee on Infomation-Based Induction Sciences and Machine Learning
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Pre-training for Action Classification Task Using Video Frame Prediction Task
Sub Title (in English)
Keyword(1) pre-training
Keyword(2) action classification
Keyword(3) video frame prediction
Keyword(4) self supervised learning
1st Author's Name Hidemoto Nakada
1st Author's Affiliation National Institute of Advanced Industrial Science and Technology(AIST)
2nd Author's Name Hideki Asoh
2nd Author's Affiliation National Institute of Advanced Industrial Science and Technology(AIST)
Date 2020-03-11
Paper # IBISML2019-45
Volume (vol) vol.119
Number (no) IBISML-476
Page pp.pp.85-90(IBISML),
#Pages 6
Date of Issue 2020-03-03 (IBISML)