Domain-Adaptive Pedestrian Detection in Thermal Images

This paper presents an approach to pedestrian detection in thermal infrared (thermal) images with limited annotations. The key idea is to adapt the abundance of color images associated with bounding box annotations to the thermal domain for training the pedestrian detector. To this end, we couple a...

Full description

Saved in:
Bibliographic Details
Main Authors: Guo, Tiantong, Huynh, Cong Phuoc, Solh, Mashhour
Format: Conference Proceeding
Language:English
Subjects:
Color
deep learning
Detectors
Generators
Image color analysis
Lighting
pedestrian detection
synthetic image
thermal image
Training
Two dimensional displays
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page 1664
container_issue
container_start_page 1660
container_title
container_volume
creator Guo, Tiantong
Huynh, Cong Phuoc
Solh, Mashhour
description This paper presents an approach to pedestrian detection in thermal infrared (thermal) images with limited annotations. The key idea is to adapt the abundance of color images associated with bounding box annotations to the thermal domain for training the pedestrian detector. To this end, we couple a domain adaptation component that consists of a pair of image transformers with a pedestrian detector in the thermal domain and train the entire network end-to-end. The image transformers act as a data augmentation tool that progressively improves synthetic examples on the fly for training the pedestrian detector. To aid the training process, we introduce a detection loss defined on both real thermal images and synthetic thermal images transformed from the color domain. The proposed detector outperforms existing methods on the thermal images from the KAIST detection benchmark [1].
doi_str_mv 10.1109/ICIP.2019.8803104
format conference_proceeding
fullrecord <record><control><sourceid>ieee_CHZPO</sourceid><recordid>TN_cdi_ieee_primary_8803104</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>8803104</ieee_id><sourcerecordid>8803104</sourcerecordid><originalsourceid>FETCH-LOGICAL-i203t-cc83ce6171233b4568e699853b483917e63583c52ac2e91de897a020c369e8743</originalsourceid><addsrcrecordid>eNotj81Kw0AUhUdBsLZ9AHEzL5A4d25mcu-ypP4ECnbRrss4uepIk5YkCL69Abs6Z_HxcY5S92ByAMOPdVVvc2uAcyKDYIorteSSwCF5bwv212pmkSAjV_CtuhuGb2MmHmGm_PrUhtRlqyacx_QjeiuNDGOfQqfXMkoc06nTqdO7L-nbcNR1Gz5lWKibj3AcZHnJudo_P-2q12zz9lJXq02WrMExi5EwiocSLOJ74TyJZyY3dUKGUjy6iXA2RCsMjRCXYZoW0bNQWeBcPfx7k4gczn1qQ_97uNzEPwOqRCU</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>Domain-Adaptive Pedestrian Detection in Thermal Images</title><source>IEEE Xplore All Conference Series</source><creator>Guo, Tiantong ; Huynh, Cong Phuoc ; Solh, Mashhour</creator><creatorcontrib>Guo, Tiantong ; Huynh, Cong Phuoc ; Solh, Mashhour</creatorcontrib><description>This paper presents an approach to pedestrian detection in thermal infrared (thermal) images with limited annotations. The key idea is to adapt the abundance of color images associated with bounding box annotations to the thermal domain for training the pedestrian detector. To this end, we couple a domain adaptation component that consists of a pair of image transformers with a pedestrian detector in the thermal domain and train the entire network end-to-end. The image transformers act as a data augmentation tool that progressively improves synthetic examples on the fly for training the pedestrian detector. To aid the training process, we introduce a detection loss defined on both real thermal images and synthetic thermal images transformed from the color domain. The proposed detector outperforms existing methods on the thermal images from the KAIST detection benchmark [1].</description><identifier>EISSN: 2381-8549</identifier><identifier>EISBN: 9781538662496</identifier><identifier>EISBN: 1538662493</identifier><identifier>DOI: 10.1109/ICIP.2019.8803104</identifier><language>eng</language><publisher>IEEE</publisher><subject>Color ; deep learning ; Detectors ; Generators ; Image color analysis ; Lighting ; pedestrian detection ; synthetic image ; thermal image ; Training ; Two dimensional displays</subject><ispartof>2019 IEEE International Conference on Image Processing (ICIP), 2019, p.1660-1664</ispartof><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/8803104$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>309,310,780,784,789,790,27925,54555,54932</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/8803104$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Guo, Tiantong</creatorcontrib><creatorcontrib>Huynh, Cong Phuoc</creatorcontrib><creatorcontrib>Solh, Mashhour</creatorcontrib><title>Domain-Adaptive Pedestrian Detection in Thermal Images</title><title>2019 IEEE International Conference on Image Processing (ICIP)</title><addtitle>ICIP</addtitle><description>This paper presents an approach to pedestrian detection in thermal infrared (thermal) images with limited annotations. The key idea is to adapt the abundance of color images associated with bounding box annotations to the thermal domain for training the pedestrian detector. To this end, we couple a domain adaptation component that consists of a pair of image transformers with a pedestrian detector in the thermal domain and train the entire network end-to-end. The image transformers act as a data augmentation tool that progressively improves synthetic examples on the fly for training the pedestrian detector. To aid the training process, we introduce a detection loss defined on both real thermal images and synthetic thermal images transformed from the color domain. The proposed detector outperforms existing methods on the thermal images from the KAIST detection benchmark [1].</description><subject>Color</subject><subject>deep learning</subject><subject>Detectors</subject><subject>Generators</subject><subject>Image color analysis</subject><subject>Lighting</subject><subject>pedestrian detection</subject><subject>synthetic image</subject><subject>thermal image</subject><subject>Training</subject><subject>Two dimensional displays</subject><issn>2381-8549</issn><isbn>9781538662496</isbn><isbn>1538662493</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2019</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNotj81Kw0AUhUdBsLZ9AHEzL5A4d25mcu-ypP4ECnbRrss4uepIk5YkCL69Abs6Z_HxcY5S92ByAMOPdVVvc2uAcyKDYIorteSSwCF5bwv212pmkSAjV_CtuhuGb2MmHmGm_PrUhtRlqyacx_QjeiuNDGOfQqfXMkoc06nTqdO7L-nbcNR1Gz5lWKibj3AcZHnJudo_P-2q12zz9lJXq02WrMExi5EwiocSLOJ74TyJZyY3dUKGUjy6iXA2RCsMjRCXYZoW0bNQWeBcPfx7k4gczn1qQ_97uNzEPwOqRCU</recordid><startdate>201909</startdate><enddate>201909</enddate><creator>Guo, Tiantong</creator><creator>Huynh, Cong Phuoc</creator><creator>Solh, Mashhour</creator><general>IEEE</general><scope>6IE</scope><scope>6IH</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIO</scope></search><sort><creationdate>201909</creationdate><title>Domain-Adaptive Pedestrian Detection in Thermal Images</title><author>Guo, Tiantong ; Huynh, Cong Phuoc ; Solh, Mashhour</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-i203t-cc83ce6171233b4568e699853b483917e63583c52ac2e91de897a020c369e8743</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Color</topic><topic>deep learning</topic><topic>Detectors</topic><topic>Generators</topic><topic>Image color analysis</topic><topic>Lighting</topic><topic>pedestrian detection</topic><topic>synthetic image</topic><topic>thermal image</topic><topic>Training</topic><topic>Two dimensional displays</topic><toplevel>online_resources</toplevel><creatorcontrib>Guo, Tiantong</creatorcontrib><creatorcontrib>Huynh, Cong Phuoc</creatorcontrib><creatorcontrib>Solh, Mashhour</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan (POP) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE Xplore (Online service)</collection><collection>IEEE Proceedings Order Plans (POP) 1998-present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Guo, Tiantong</au><au>Huynh, Cong Phuoc</au><au>Solh, Mashhour</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>Domain-Adaptive Pedestrian Detection in Thermal Images</atitle><btitle>2019 IEEE International Conference on Image Processing (ICIP)</btitle><stitle>ICIP</stitle><date>2019-09</date><risdate>2019</risdate><spage>1660</spage><epage>1664</epage><pages>1660-1664</pages><eissn>2381-8549</eissn><eisbn>9781538662496</eisbn><eisbn>1538662493</eisbn><abstract>This paper presents an approach to pedestrian detection in thermal infrared (thermal) images with limited annotations. The key idea is to adapt the abundance of color images associated with bounding box annotations to the thermal domain for training the pedestrian detector. To this end, we couple a domain adaptation component that consists of a pair of image transformers with a pedestrian detector in the thermal domain and train the entire network end-to-end. The image transformers act as a data augmentation tool that progressively improves synthetic examples on the fly for training the pedestrian detector. To aid the training process, we introduce a detection loss defined on both real thermal images and synthetic thermal images transformed from the color domain. The proposed detector outperforms existing methods on the thermal images from the KAIST detection benchmark [1].</abstract><pub>IEEE</pub><doi>10.1109/ICIP.2019.8803104</doi><tpages>5</tpages></addata></record>
fulltext fulltext_linktorsrc
identifier EISSN: 2381-8549
ispartof 2019 IEEE International Conference on Image Processing (ICIP), 2019, p.1660-1664
issn 2381-8549
language eng
recordid cdi_ieee_primary_8803104
source IEEE Xplore All Conference Series
subjects Color
deep learning
Detectors
Generators
Image color analysis
Lighting
pedestrian detection
synthetic image
thermal image
Training
Two dimensional displays
title Domain-Adaptive Pedestrian Detection in Thermal Images
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T02%3A48%3A03IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_CHZPO&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=Domain-Adaptive%20Pedestrian%20Detection%20in%20Thermal%20Images&rft.btitle=2019%20IEEE%20International%20Conference%20on%20Image%20Processing%20(ICIP)&rft.au=Guo,%20Tiantong&rft.date=2019-09&rft.spage=1660&rft.epage=1664&rft.pages=1660-1664&rft.eissn=2381-8549&rft_id=info:doi/10.1109/ICIP.2019.8803104&rft.eisbn=9781538662496&rft.eisbn_list=1538662493&rft_dat=%3Cieee_CHZPO%3E8803104%3C/ieee_CHZPO%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-i203t-cc83ce6171233b4568e699853b483917e63583c52ac2e91de897a020c369e8743%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=8803104&rfr_iscdi=true