Personness estimation for real-time human detection on mobile devices

Kyuwon Kim, Changjae Oh, Kwanghoon Sohn

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

One aim of detection proposal methods is to reduce the computational overhead of object detection. However, most of the existing methods have significant computational overhead for real-time detection on mobile devices. A fast and accurate proposal method of human detection called personness estimation is proposed, which facilitates real-time human detection on mobile devices and can be effectively integrated into part-based detection, achieving high detection performance at a low computational cost. Our work is based on two observations: (i) normed gradients, which are designed for generic objectness estimation, effectively generate high-quality detection proposals for the person category; (ii) fusing the normed gradients with color attributes improves the performance of proposal generation for human detection. Thus, the candidate windows generated by the personness estimation will very likely contain human subjects. The human detection is then guided by the candidate windows, offering high detection performance even when the detection task terminates prior to completion. This interruptible detection scheme, called anytime detection, enables real-time human detection on mobile devices. Furthermore, we introduce a new evaluation methodology called time-recall curves to practically evaluate our approach. The applicability of our proposed method is demonstrated in extensive experiments on a publicly available dataset and a real mobile device, facilitating acquisition and enhancement of portrait photographs (e.g. selfie) on widespread mobile platforms.

Original languageEnglish
Pages (from-to)130-138
Number of pages9
JournalExpert Systems with Applications
Volume72
DOIs
Publication statusPublished - 2017 Apr 15

Fingerprint

Mobile devices
Color
Costs
Experiments

All Science Journal Classification (ASJC) codes

  • Engineering(all)
  • Computer Science Applications
  • Artificial Intelligence

Cite this

@article{d6eb4267308f40cfb9ab7c7a1ccbb340,
title = "Personness estimation for real-time human detection on mobile devices",
abstract = "One aim of detection proposal methods is to reduce the computational overhead of object detection. However, most of the existing methods have significant computational overhead for real-time detection on mobile devices. A fast and accurate proposal method of human detection called personness estimation is proposed, which facilitates real-time human detection on mobile devices and can be effectively integrated into part-based detection, achieving high detection performance at a low computational cost. Our work is based on two observations: (i) normed gradients, which are designed for generic objectness estimation, effectively generate high-quality detection proposals for the person category; (ii) fusing the normed gradients with color attributes improves the performance of proposal generation for human detection. Thus, the candidate windows generated by the personness estimation will very likely contain human subjects. The human detection is then guided by the candidate windows, offering high detection performance even when the detection task terminates prior to completion. This interruptible detection scheme, called anytime detection, enables real-time human detection on mobile devices. Furthermore, we introduce a new evaluation methodology called time-recall curves to practically evaluate our approach. The applicability of our proposed method is demonstrated in extensive experiments on a publicly available dataset and a real mobile device, facilitating acquisition and enhancement of portrait photographs (e.g. selfie) on widespread mobile platforms.",
author = "Kyuwon Kim and Changjae Oh and Kwanghoon Sohn",
year = "2017",
month = "4",
day = "15",
doi = "10.1016/j.eswa.2016.12.017",
language = "English",
volume = "72",
pages = "130--138",
journal = "Expert Systems with Applications",
issn = "0957-4174",
publisher = "Elsevier Limited",

}

Personness estimation for real-time human detection on mobile devices. / Kim, Kyuwon; Oh, Changjae; Sohn, Kwanghoon.

In: Expert Systems with Applications, Vol. 72, 15.04.2017, p. 130-138.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Personness estimation for real-time human detection on mobile devices

AU - Kim, Kyuwon

AU - Oh, Changjae

AU - Sohn, Kwanghoon

PY - 2017/4/15

Y1 - 2017/4/15

N2 - One aim of detection proposal methods is to reduce the computational overhead of object detection. However, most of the existing methods have significant computational overhead for real-time detection on mobile devices. A fast and accurate proposal method of human detection called personness estimation is proposed, which facilitates real-time human detection on mobile devices and can be effectively integrated into part-based detection, achieving high detection performance at a low computational cost. Our work is based on two observations: (i) normed gradients, which are designed for generic objectness estimation, effectively generate high-quality detection proposals for the person category; (ii) fusing the normed gradients with color attributes improves the performance of proposal generation for human detection. Thus, the candidate windows generated by the personness estimation will very likely contain human subjects. The human detection is then guided by the candidate windows, offering high detection performance even when the detection task terminates prior to completion. This interruptible detection scheme, called anytime detection, enables real-time human detection on mobile devices. Furthermore, we introduce a new evaluation methodology called time-recall curves to practically evaluate our approach. The applicability of our proposed method is demonstrated in extensive experiments on a publicly available dataset and a real mobile device, facilitating acquisition and enhancement of portrait photographs (e.g. selfie) on widespread mobile platforms.

AB - One aim of detection proposal methods is to reduce the computational overhead of object detection. However, most of the existing methods have significant computational overhead for real-time detection on mobile devices. A fast and accurate proposal method of human detection called personness estimation is proposed, which facilitates real-time human detection on mobile devices and can be effectively integrated into part-based detection, achieving high detection performance at a low computational cost. Our work is based on two observations: (i) normed gradients, which are designed for generic objectness estimation, effectively generate high-quality detection proposals for the person category; (ii) fusing the normed gradients with color attributes improves the performance of proposal generation for human detection. Thus, the candidate windows generated by the personness estimation will very likely contain human subjects. The human detection is then guided by the candidate windows, offering high detection performance even when the detection task terminates prior to completion. This interruptible detection scheme, called anytime detection, enables real-time human detection on mobile devices. Furthermore, we introduce a new evaluation methodology called time-recall curves to practically evaluate our approach. The applicability of our proposed method is demonstrated in extensive experiments on a publicly available dataset and a real mobile device, facilitating acquisition and enhancement of portrait photographs (e.g. selfie) on widespread mobile platforms.

UR - http://www.scopus.com/inward/record.url?scp=85007552063&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85007552063&partnerID=8YFLogxK

U2 - 10.1016/j.eswa.2016.12.017

DO - 10.1016/j.eswa.2016.12.017

M3 - Article

AN - SCOPUS:85007552063

VL - 72

SP - 130

EP - 138

JO - Expert Systems with Applications

JF - Expert Systems with Applications

SN - 0957-4174

ER -