Presentation 2020-05-14
Human Action Recognition with Two-stream 3D BagNet
Junya Uchida, Yu Wang, Jien Kato,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose the Two-stream 3D BagNet for the human action recognition task. The proposed architecture is inspired by the concepts of Two-stream I3D and BagNet, and enjoys both good accuracy and interpretability. Specifically, we designed 3D BagNet 9, 17 and 33 based on 3D ResNet18. The receptive fields of these models are enforced to be small by eliminating padding and replacing most 3*3*3 filters with 1*1*1 ones. The spatial version and temporal version of these BagNets were trained using RGB images and optical flow images respectively. We evaluated these models’ performances and interpretabilities extensively. We also confirmed by fusing the spatial and temporal nets, the recognition performance can be significantly improved without sacrificing the interpretability.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Action recognition / DNN / 3D CNN / Local features / Grounds for recognition
Paper # PRMU2020-3
Date of Issue 2020-05-07 (PRMU)

Conference Information
Committee PRMU / IPSJ-CVIM
Conference Date 2020/5/14(2days)
Place (in Japanese) (See Japanese page)
Place (in English) online
Topics (in Japanese) (See Japanese page)
Topics (in English) Business based on computer vision and pattern recognition
Chair Yoichi Sato(Univ. of Tokyo)
Vice Chair Toru Tamaki(Hiroshima Univ.) / Akisato Kimura(NTT)
Secretary Toru Tamaki(NTT) / Akisato Kimura(OMRON SINICX)
Assistant Yusuke Uchida(DeNA) / Takayoshi Yamashita(Chubu Univ.)

Paper Information
Registration To Technical Committee on Pattern Recognition and Media Understanding / Special Interest Group on Computer Vision and Image Media
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Human Action Recognition with Two-stream 3D BagNet
Sub Title (in English)
Keyword(1) Action recognition
Keyword(2) DNN
Keyword(3) 3D CNN
Keyword(4) Local features
Keyword(5) Grounds for recognition
1st Author's Name Junya Uchida
1st Author's Affiliation Ritsumeikan University(Ritsumeikan Univ.)
2nd Author's Name Yu Wang
2nd Author's Affiliation Ritsumeikan University(Ritsumeikan Univ.)
3rd Author's Name Jien Kato
3rd Author's Affiliation Ritsumeikan University(Ritsumeikan Univ.)
Date 2020-05-14
Paper # PRMU2020-3
Volume (vol) vol.120
Number (no) PRMU-14
Page pp.pp.13-17(PRMU),
#Pages 5
Date of Issue 2020-05-07 (PRMU)