Publication detail

Real-Time Pose Estimation Piggybacked on Object Detection

JURÁNEK, R. HEROUT, A. JURÁNKOVÁ, M. ZEMČÍK, P.

Original Title

Real-Time Pose Estimation Piggybacked on Object Detection

English Title

Real-Time Pose Estimation Piggybacked on Object Detection

Type

conference paper

Language

en

Original Abstract

We present an object detector coupled with pose estimation directly in a single compact and simple model, where the detector shares extracted image features with the pose estimator. The output of the classification of each candidate window consists of both object score and likelihood map of poses. This extension introduces negligible overhead during detection so that the detector is still capable of real time operation. We evaluated the proposed approach on the problem of vehicle detection. We used existing datasets with viewpoint/pose annotation (WCVP, 3D objects, KITTI).  Besides that, we collected a new traffic surveillance dataset COD20k which fills certain gaps of the existing datasets and we make it public. The experimental results show that the proposed approach is comparable with state-of-the-art approaches in terms of accuracy, but it is considerably faster -- easily operating in real time (Matlab with C++ code). The source codes and the collected COD20k dataset are made public along with the paper.

English abstract

We present an object detector coupled with pose estimation directly in a single compact and simple model, where the detector shares extracted image features with the pose estimator. The output of the classification of each candidate window consists of both object score and likelihood map of poses. This extension introduces negligible overhead during detection so that the detector is still capable of real time operation. We evaluated the proposed approach on the problem of vehicle detection. We used existing datasets with viewpoint/pose annotation (WCVP, 3D objects, KITTI).  Besides that, we collected a new traffic surveillance dataset COD20k which fills certain gaps of the existing datasets and we make it public. The experimental results show that the proposed approach is comparable with state-of-the-art approaches in terms of accuracy, but it is considerably faster -- easily operating in real time (Matlab with C++ code). The source codes and the collected COD20k dataset are made public along with the paper.

Keywords

Object detection, Pose estimation, Sliding window detector, Channel features

RIV year

2015

Released

18.12.2015

Publisher

IEEE Computer Society

Location

Santiago

ISBN

978-1-4673-8391-2

Book

Proceedings of ICCV

Edition

NEUVEDEN

Edition number

NEUVEDEN

Pages from

1

Pages to

9

Pages count

9

Documents

BibTex


@inproceedings{BUT123622,
  author="Roman {Juránek} and Adam {Herout} and Markéta {Juránková} and Pavel {Zemčík}",
  title="Real-Time Pose Estimation Piggybacked on Object Detection",
  annote="We present an object detector coupled with pose estimation directly in a single
compact and simple model, where the detector shares extracted image features with
the pose estimator. The output of the classification of each candidate window
consists of both object score and likelihood map of poses. This extension
introduces negligible overhead during detection so that the detector is still
capable of real time operation. We evaluated the proposed approach on the problem
of vehicle detection. We used existing datasets with viewpoint/pose annotation
(WCVP, 3D objects, KITTI).  Besides that, we collected a new traffic surveillance
dataset COD20k which fills certain gaps of the existing datasets and we make it
public. The experimental results show that the proposed approach is comparable
with state-of-the-art approaches in terms of accuracy, but it is considerably
faster -- easily operating in real time (Matlab with C++ code). The source codes
and the collected COD20k dataset are made public along with the paper.",
  address="IEEE Computer Society",
  booktitle="Proceedings of ICCV",
  chapter="123622",
  edition="NEUVEDEN",
  howpublished="online",
  institution="IEEE Computer Society",
  year="2015",
  month="december",
  pages="1--9",
  publisher="IEEE Computer Society",
  type="conference paper"
}