yiskw713 / paper_summary

Write summaries of papers I've read in Japanese
10 stars 1 forks source link

Human Action Recognition without Human #113

Open yiskw713 opened 4 years ago

yiskw713 commented 4 years ago

INFO

author

Yun He, Soma Shirakabe, Yutaka Satoh, Hirokatsu Kataoka

affiliation

conference or year

ECCV2016, workshop

link

pdf 解説スライド

概要

Screenshot from 2020-01-17 16-49-41

行動認識において,背景情報だけでも行動認識ができてしまうことを指摘した論文. two-stream network, UCF-101を用いて,動画の中心領域(with humans)で学習させた場合と,外側の領域(without humans)で学習させた場合には,精度の差は9%ほどしかなかった.

Comment

date

Jan. 17th, 2020