Unsupervised pedestrian sample extraction for model training

Yu Hao, Daryl Marples, Ying Liu, Zhijie Xu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many researches on pedestrian detection use benchmarking datasets such as INRIA for model training. However, models trained with standard video database do not usually obtain satisfying performance in real-life conditions. Hence, supervised training through manually labelled instances is often required to help achieving better detection result. In this research, an innovative unsupervised training approach is proposed. By analyzing the histogram of adjacent pixels modelled from the video sequences, separated pedestrians can be extracted without manual intervention. Experiments have shown consistent performance that is superior over the state-of-the-art methods.
LanguageEnglish
Title of host publicationProceedings of the International Conferences
Subtitle of host publication Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019
EditorsKatherine Blashki, Yingcai Xiao
PublisherIADIS
Pages291-298
Number of pages8
ISBN (Electronic)9789898533913
Publication statusPublished - 16 Jul 2019
Event13th Multi Conference on Computer Science and Information Systems 2019: IADIS International Conferences Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019 - Porto, Portugal
Duration: 16 Jul 201919 Jul 2019
http://www.iadisportal.org/digital-library/iadis-international-conference-game-and-entertainment-technologies-2019-part-of-mccsis-2019

Publication series

NameMulti conference on computer science and information systems 2019
PublisherIADIS

Conference

Conference13th Multi Conference on Computer Science and Information Systems 2019
Abbreviated titleMCCSIS 2019
CountryPortugal
CityPorto
Period16/07/1919/07/19
Internet address

Fingerprint

Benchmarking
Pixels
Experiments

Cite this

Hao, Y., Marples, D., Liu, Y., & Xu, Z. (2019). Unsupervised pedestrian sample extraction for model training. In K. Blashki, & Y. Xiao (Eds.), Proceedings of the International Conferences: Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019 (pp. 291-298). (Multi conference on computer science and information systems 2019). IADIS.
Hao, Yu ; Marples, Daryl ; Liu, Ying ; Xu, Zhijie. / Unsupervised pedestrian sample extraction for model training. Proceedings of the International Conferences: Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019. editor / Katherine Blashki ; Yingcai Xiao. IADIS, 2019. pp. 291-298 (Multi conference on computer science and information systems 2019).
@inproceedings{0f3361752e1547ec905e1175d874d9c8,
title = "Unsupervised pedestrian sample extraction for model training",
abstract = "Many researches on pedestrian detection use benchmarking datasets such as INRIA for model training. However, models trained with standard video database do not usually obtain satisfying performance in real-life conditions. Hence, supervised training through manually labelled instances is often required to help achieving better detection result. In this research, an innovative unsupervised training approach is proposed. By analyzing the histogram of adjacent pixels modelled from the video sequences, separated pedestrians can be extracted without manual intervention. Experiments have shown consistent performance that is superior over the state-of-the-art methods.",
keywords = "Pedestrian Detection, Crowd Analysis, Unsupervised Learning",
author = "Yu Hao and Daryl Marples and Ying Liu and Zhijie Xu",
year = "2019",
month = "7",
day = "16",
language = "English",
series = "Multi conference on computer science and information systems 2019",
publisher = "IADIS",
pages = "291--298",
editor = "Katherine Blashki and Yingcai Xiao",
booktitle = "Proceedings of the International Conferences",

}

Hao, Y, Marples, D, Liu, Y & Xu, Z 2019, Unsupervised pedestrian sample extraction for model training. in K Blashki & Y Xiao (eds), Proceedings of the International Conferences: Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019. Multi conference on computer science and information systems 2019, IADIS, pp. 291-298, 13th Multi Conference on Computer Science and Information Systems 2019, Porto, Portugal, 16/07/19.

Unsupervised pedestrian sample extraction for model training. / Hao, Yu; Marples, Daryl; Liu, Ying; Xu, Zhijie.

Proceedings of the International Conferences: Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019. ed. / Katherine Blashki; Yingcai Xiao. IADIS, 2019. p. 291-298 (Multi conference on computer science and information systems 2019).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Unsupervised pedestrian sample extraction for model training

AU - Hao, Yu

AU - Marples, Daryl

AU - Liu, Ying

AU - Xu, Zhijie

PY - 2019/7/16

Y1 - 2019/7/16

N2 - Many researches on pedestrian detection use benchmarking datasets such as INRIA for model training. However, models trained with standard video database do not usually obtain satisfying performance in real-life conditions. Hence, supervised training through manually labelled instances is often required to help achieving better detection result. In this research, an innovative unsupervised training approach is proposed. By analyzing the histogram of adjacent pixels modelled from the video sequences, separated pedestrians can be extracted without manual intervention. Experiments have shown consistent performance that is superior over the state-of-the-art methods.

AB - Many researches on pedestrian detection use benchmarking datasets such as INRIA for model training. However, models trained with standard video database do not usually obtain satisfying performance in real-life conditions. Hence, supervised training through manually labelled instances is often required to help achieving better detection result. In this research, an innovative unsupervised training approach is proposed. By analyzing the histogram of adjacent pixels modelled from the video sequences, separated pedestrians can be extracted without manual intervention. Experiments have shown consistent performance that is superior over the state-of-the-art methods.

KW - Pedestrian Detection

KW - Crowd Analysis

KW - Unsupervised Learning

UR - http://www.iadisportal.org/digital-library/iadis-international-conference-game-and-entertainment-technologies-2019-part-of-mccsis-2019

M3 - Conference contribution

T3 - Multi conference on computer science and information systems 2019

SP - 291

EP - 298

BT - Proceedings of the International Conferences

A2 - Blashki, Katherine

A2 - Xiao, Yingcai

PB - IADIS

ER -

Hao Y, Marples D, Liu Y, Xu Z. Unsupervised pedestrian sample extraction for model training. In Blashki K, Xiao Y, editors, Proceedings of the International Conferences: Interfaces and Human Computer Interaction 2019; Game and Entertainment Technologies 2019; and Computer Graphics, Visualization, Computer Vision and Image Processing 2019. IADIS. 2019. p. 291-298. (Multi conference on computer science and information systems 2019).