Ensemble convolutional neural networks for pose estimation

Yuki Kawana, Norimichi Ukita, Jia Bin Huang, Ming Hsuan Yang

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Human pose estimation is a challenging task due to significant appearance variations. An ensemble of models, each of which is optimized for a limited variety of poses, is capable of modeling a large variety of human body configurations.However, ensembling models is not a straightforward task due to the complex interdependence among noisy and ambiguous pose estimation predictions acquired by each model.We propose to capture this complex interdependence using a convolutional neural network. Our network achieves this interdependence representation using a combination of deep convolution and deconvolution layers for robust and accurate pose estimation. We evaluate the proposed ensemble model on publicly available datasets and show that our model compares favorably against baseline models and state-of-the-art methods.

Original languageEnglish
Pages (from-to)62-74
Number of pages13
JournalComputer Vision and Image Understanding
Volume169
DOIs
Publication statusPublished - 2018 Apr

Fingerprint

Neural networks
Deconvolution
Convolution

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition

Cite this

Kawana, Yuki ; Ukita, Norimichi ; Huang, Jia Bin ; Yang, Ming Hsuan. / Ensemble convolutional neural networks for pose estimation. In: Computer Vision and Image Understanding. 2018 ; Vol. 169. pp. 62-74.
@article{a602637c94b04fb1ae99ad5ef69b98d7,
title = "Ensemble convolutional neural networks for pose estimation",
abstract = "Human pose estimation is a challenging task due to significant appearance variations. An ensemble of models, each of which is optimized for a limited variety of poses, is capable of modeling a large variety of human body configurations.However, ensembling models is not a straightforward task due to the complex interdependence among noisy and ambiguous pose estimation predictions acquired by each model.We propose to capture this complex interdependence using a convolutional neural network. Our network achieves this interdependence representation using a combination of deep convolution and deconvolution layers for robust and accurate pose estimation. We evaluate the proposed ensemble model on publicly available datasets and show that our model compares favorably against baseline models and state-of-the-art methods.",
author = "Yuki Kawana and Norimichi Ukita and Huang, {Jia Bin} and Yang, {Ming Hsuan}",
year = "2018",
month = "4",
doi = "10.1016/j.cviu.2017.12.005",
language = "English",
volume = "169",
pages = "62--74",
journal = "Computer Vision and Image Understanding",
issn = "1077-3142",
publisher = "Academic Press Inc.",

}

Ensemble convolutional neural networks for pose estimation. / Kawana, Yuki; Ukita, Norimichi; Huang, Jia Bin; Yang, Ming Hsuan.

In: Computer Vision and Image Understanding, Vol. 169, 04.2018, p. 62-74.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Ensemble convolutional neural networks for pose estimation

AU - Kawana, Yuki

AU - Ukita, Norimichi

AU - Huang, Jia Bin

AU - Yang, Ming Hsuan

PY - 2018/4

Y1 - 2018/4

N2 - Human pose estimation is a challenging task due to significant appearance variations. An ensemble of models, each of which is optimized for a limited variety of poses, is capable of modeling a large variety of human body configurations.However, ensembling models is not a straightforward task due to the complex interdependence among noisy and ambiguous pose estimation predictions acquired by each model.We propose to capture this complex interdependence using a convolutional neural network. Our network achieves this interdependence representation using a combination of deep convolution and deconvolution layers for robust and accurate pose estimation. We evaluate the proposed ensemble model on publicly available datasets and show that our model compares favorably against baseline models and state-of-the-art methods.

AB - Human pose estimation is a challenging task due to significant appearance variations. An ensemble of models, each of which is optimized for a limited variety of poses, is capable of modeling a large variety of human body configurations.However, ensembling models is not a straightforward task due to the complex interdependence among noisy and ambiguous pose estimation predictions acquired by each model.We propose to capture this complex interdependence using a convolutional neural network. Our network achieves this interdependence representation using a combination of deep convolution and deconvolution layers for robust and accurate pose estimation. We evaluate the proposed ensemble model on publicly available datasets and show that our model compares favorably against baseline models and state-of-the-art methods.

UR - http://www.scopus.com/inward/record.url?scp=85040336059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040336059&partnerID=8YFLogxK

U2 - 10.1016/j.cviu.2017.12.005

DO - 10.1016/j.cviu.2017.12.005

M3 - Article

AN - SCOPUS:85040336059

VL - 169

SP - 62

EP - 74

JO - Computer Vision and Image Understanding

JF - Computer Vision and Image Understanding

SN - 1077-3142

ER -