Paper published in a book (Scientific congresses, symposiums and conference proceedings)
Temporal 3D Human Pose Estimation for Action Recognition from Arbitrary Viewpoints
Adel Musallam, Mohamed; BAPTISTA, Renato; AL ISMAEIL, Kassem et al.
2019In 6th Annual Conf. on Computational Science & Computational Intelligence, Las Vegas 5-7 December 2019
Peer reviewed
 

Files


Full Text
csci_cameraready_2019.pdf
Author postprint (6.01 MB)
Download

All documents in ORBilu are protected by a user license.

Send to



Details



Keywords :
View-Invariant; Human Action Recognition; Human Pose Estimation
Abstract :
[en] This work presents a new view-invariant action recognition system that is able to classify human actions by using a single RGB camera, including challenging camera viewpoints. Understanding actions from different viewpoints remains an extremely challenging problem, due to depth ambiguities, occlusion, and a large variety of appearances and scenes. Moreover, using only the information from the 2D perspective gives different interpretations for the same action seen from different viewpoints. Our system operates in two subsequent stages. The first stage estimates the 2D human pose using a convolution neural network. In the next stage, the 2D human poses are lifted to 3D human poses, using a temporal convolution neural network that enforces the temporal coherence over the estimated 3D poses. The estimated 3D poses from different viewpoints are then aligned to the same camera reference frame. Finally, we propose to use a temporal convolution network-based classifier for cross-view action recognition. Our results show that we can achieve state of art view-invariant action recognition accuracy even for the challenging viewpoints by only using RGB videos, without pre-training on synthetic or motion capture data.
Research center :
Interdisciplinary Centre for Security, Reliability and Trust (SnT) > SIGCOM
Disciplines :
Computer science
Author, co-author :
Adel Musallam, Mohamed
BAPTISTA, Renato ;  University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT)
AL ISMAEIL, Kassem ;  University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT)
AOUADA, Djamila  ;  University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT)
External co-authors :
no
Language :
English
Title :
Temporal 3D Human Pose Estimation for Action Recognition from Arbitrary Viewpoints
Publication date :
December 2019
Event name :
6th Annual Conf. on Computational Science & Computational Intelligence
Event organizer :
https://americancse.org/events/csci2019
Event date :
5-7 December 2019
Audience :
International
Main work title :
6th Annual Conf. on Computational Science & Computational Intelligence, Las Vegas 5-7 December 2019
Publisher :
Conference Publishing Services
Peer reviewed :
Peer reviewed
Focus Area :
Computational Sciences
European Projects :
H2020 - 689947 - STARR - Decision SupporT and self-mAnagement system for stRoke survivoRs
FnR Project :
FNR10415355 - 3d Action Recognition Using Refinement And Invariance Strategies For Reliable Surveillance, 2015 (01/06/2016-31/05/2019) - Bjorn Ottersten
Funders :
CE - Commission Européenne [BE]
Available on ORBilu :
since 30 November 2019

Statistics


Number of views
351 (16 by Unilu)
Number of downloads
407 (9 by Unilu)

Scopus citations®
 
3
Scopus citations®
without self-citations
2

Bibliography


Similar publications



Contact ORBilu