CLAD: A realistic Continual Learning benchmark for Autonomous Driving

In this paper we describe the design and the ideas motivating a new Continual Learning benchmark for Autonomous Driving (CLAD), that focuses on the problems of object classification and object detection. The benchmark utilises SODA10M, a recently released large-scale dataset that concerns autonomous...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:NEURAL NETWORKS 2023-04, Vol.161, p.659-669
Hauptverfasser: Verwimp, Eli, Yang, Kuo, Parisot, Sarah, Hong, Lanqing, McDonagh, Steven, Perez-Pellitero, Eduardo, De Lange, Matthias, Tuytelaars, Tinne
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 669
container_issue
container_start_page 659
container_title NEURAL NETWORKS
container_volume 161
creator Verwimp, Eli
Yang, Kuo
Parisot, Sarah
Hong, Lanqing
McDonagh, Steven
Perez-Pellitero, Eduardo
De Lange, Matthias
Tuytelaars, Tinne
description In this paper we describe the design and the ideas motivating a new Continual Learning benchmark for Autonomous Driving (CLAD), that focuses on the problems of object classification and object detection. The benchmark utilises SODA10M, a recently released large-scale dataset that concerns autonomous driving related problems. First, we review and discuss existing continual learning benchmarks, how they are related, and show that most are extreme cases of continual learning. To this end, we survey the benchmarks used in continual learning papers at three highly ranked computer vision conferences. Next, we introduce CLAD-C, an online classification benchmark realised through a chronological data stream that poses both class and domain incremental challenges; and CLAD-D, a domain incremental continual object detection benchmark. We examine the inherent difficulties and challenges posed by the benchmark, through a survey of the techniques and methods used by the top-3 participants in a CLAD-challenge workshop at ICCV 2021. We conclude with possible pathways to improve the current continual learning state of the art, and which directions we deem promising for future research.
format Article
fullrecord <record><control><sourceid>kuleuven</sourceid><recordid>TN_cdi_kuleuven_dspace_20_500_12942_735519</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>20_500_12942_735519</sourcerecordid><originalsourceid>FETCH-kuleuven_dspace_20_500_12942_7355193</originalsourceid><addsrcrecordid>eNqVyrsOgjAUANAOmoiPf-hsgrlQQHAjgHFgdG8qFq2U1vRB_Hwd_ACdznJmKIC8IGEGOSzQ0toHAGR5QgLUVG1ZH3CJDWdSWCc6XGnlhPJM4pYzo4S64QtX3X1kZsC9Nrj0Tis9am9xbcT0CWs075m0fPN1hbbH5lydwsFL7ieu6NU-WcdpDDQFoFFcJDHdkzSNCvJn3v2cqXs58gafV0rp</addsrcrecordid><sourcetype>Institutional Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>CLAD: A realistic Continual Learning benchmark for Autonomous Driving</title><source>Lirias (KU Leuven Association)</source><source>Access via ScienceDirect (Elsevier)</source><creator>Verwimp, Eli ; Yang, Kuo ; Parisot, Sarah ; Hong, Lanqing ; McDonagh, Steven ; Perez-Pellitero, Eduardo ; De Lange, Matthias ; Tuytelaars, Tinne</creator><creatorcontrib>Verwimp, Eli ; Yang, Kuo ; Parisot, Sarah ; Hong, Lanqing ; McDonagh, Steven ; Perez-Pellitero, Eduardo ; De Lange, Matthias ; Tuytelaars, Tinne</creatorcontrib><description>In this paper we describe the design and the ideas motivating a new Continual Learning benchmark for Autonomous Driving (CLAD), that focuses on the problems of object classification and object detection. The benchmark utilises SODA10M, a recently released large-scale dataset that concerns autonomous driving related problems. First, we review and discuss existing continual learning benchmarks, how they are related, and show that most are extreme cases of continual learning. To this end, we survey the benchmarks used in continual learning papers at three highly ranked computer vision conferences. Next, we introduce CLAD-C, an online classification benchmark realised through a chronological data stream that poses both class and domain incremental challenges; and CLAD-D, a domain incremental continual object detection benchmark. We examine the inherent difficulties and challenges posed by the benchmark, through a survey of the techniques and methods used by the top-3 participants in a CLAD-challenge workshop at ICCV 2021. We conclude with possible pathways to improve the current continual learning state of the art, and which directions we deem promising for future research.</description><identifier>ISSN: 0893-6080</identifier><language>eng</language><publisher>PERGAMON-ELSEVIER SCIENCE LTD</publisher><ispartof>NEURAL NETWORKS, 2023-04, Vol.161, p.659-669</ispartof><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,315,780,784,27860</link.rule.ids></links><search><creatorcontrib>Verwimp, Eli</creatorcontrib><creatorcontrib>Yang, Kuo</creatorcontrib><creatorcontrib>Parisot, Sarah</creatorcontrib><creatorcontrib>Hong, Lanqing</creatorcontrib><creatorcontrib>McDonagh, Steven</creatorcontrib><creatorcontrib>Perez-Pellitero, Eduardo</creatorcontrib><creatorcontrib>De Lange, Matthias</creatorcontrib><creatorcontrib>Tuytelaars, Tinne</creatorcontrib><title>CLAD: A realistic Continual Learning benchmark for Autonomous Driving</title><title>NEURAL NETWORKS</title><description>In this paper we describe the design and the ideas motivating a new Continual Learning benchmark for Autonomous Driving (CLAD), that focuses on the problems of object classification and object detection. The benchmark utilises SODA10M, a recently released large-scale dataset that concerns autonomous driving related problems. First, we review and discuss existing continual learning benchmarks, how they are related, and show that most are extreme cases of continual learning. To this end, we survey the benchmarks used in continual learning papers at three highly ranked computer vision conferences. Next, we introduce CLAD-C, an online classification benchmark realised through a chronological data stream that poses both class and domain incremental challenges; and CLAD-D, a domain incremental continual object detection benchmark. We examine the inherent difficulties and challenges posed by the benchmark, through a survey of the techniques and methods used by the top-3 participants in a CLAD-challenge workshop at ICCV 2021. We conclude with possible pathways to improve the current continual learning state of the art, and which directions we deem promising for future research.</description><issn>0893-6080</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>FZOIL</sourceid><recordid>eNqVyrsOgjAUANAOmoiPf-hsgrlQQHAjgHFgdG8qFq2U1vRB_Hwd_ACdznJmKIC8IGEGOSzQ0toHAGR5QgLUVG1ZH3CJDWdSWCc6XGnlhPJM4pYzo4S64QtX3X1kZsC9Nrj0Tis9am9xbcT0CWs075m0fPN1hbbH5lydwsFL7ieu6NU-WcdpDDQFoFFcJDHdkzSNCvJn3v2cqXs58gafV0rp</recordid><startdate>202304</startdate><enddate>202304</enddate><creator>Verwimp, Eli</creator><creator>Yang, Kuo</creator><creator>Parisot, Sarah</creator><creator>Hong, Lanqing</creator><creator>McDonagh, Steven</creator><creator>Perez-Pellitero, Eduardo</creator><creator>De Lange, Matthias</creator><creator>Tuytelaars, Tinne</creator><general>PERGAMON-ELSEVIER SCIENCE LTD</general><scope>FZOIL</scope></search><sort><creationdate>202304</creationdate><title>CLAD: A realistic Continual Learning benchmark for Autonomous Driving</title><author>Verwimp, Eli ; Yang, Kuo ; Parisot, Sarah ; Hong, Lanqing ; McDonagh, Steven ; Perez-Pellitero, Eduardo ; De Lange, Matthias ; Tuytelaars, Tinne</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-kuleuven_dspace_20_500_12942_7355193</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Verwimp, Eli</creatorcontrib><creatorcontrib>Yang, Kuo</creatorcontrib><creatorcontrib>Parisot, Sarah</creatorcontrib><creatorcontrib>Hong, Lanqing</creatorcontrib><creatorcontrib>McDonagh, Steven</creatorcontrib><creatorcontrib>Perez-Pellitero, Eduardo</creatorcontrib><creatorcontrib>De Lange, Matthias</creatorcontrib><creatorcontrib>Tuytelaars, Tinne</creatorcontrib><collection>Lirias (KU Leuven Association)</collection><jtitle>NEURAL NETWORKS</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Verwimp, Eli</au><au>Yang, Kuo</au><au>Parisot, Sarah</au><au>Hong, Lanqing</au><au>McDonagh, Steven</au><au>Perez-Pellitero, Eduardo</au><au>De Lange, Matthias</au><au>Tuytelaars, Tinne</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>CLAD: A realistic Continual Learning benchmark for Autonomous Driving</atitle><jtitle>NEURAL NETWORKS</jtitle><date>2023-04</date><risdate>2023</risdate><volume>161</volume><spage>659</spage><epage>669</epage><pages>659-669</pages><issn>0893-6080</issn><abstract>In this paper we describe the design and the ideas motivating a new Continual Learning benchmark for Autonomous Driving (CLAD), that focuses on the problems of object classification and object detection. The benchmark utilises SODA10M, a recently released large-scale dataset that concerns autonomous driving related problems. First, we review and discuss existing continual learning benchmarks, how they are related, and show that most are extreme cases of continual learning. To this end, we survey the benchmarks used in continual learning papers at three highly ranked computer vision conferences. Next, we introduce CLAD-C, an online classification benchmark realised through a chronological data stream that poses both class and domain incremental challenges; and CLAD-D, a domain incremental continual object detection benchmark. We examine the inherent difficulties and challenges posed by the benchmark, through a survey of the techniques and methods used by the top-3 participants in a CLAD-challenge workshop at ICCV 2021. We conclude with possible pathways to improve the current continual learning state of the art, and which directions we deem promising for future research.</abstract><pub>PERGAMON-ELSEVIER SCIENCE LTD</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 0893-6080
ispartof NEURAL NETWORKS, 2023-04, Vol.161, p.659-669
issn 0893-6080
language eng
recordid cdi_kuleuven_dspace_20_500_12942_735519
source Lirias (KU Leuven Association); Access via ScienceDirect (Elsevier)
title CLAD: A realistic Continual Learning benchmark for Autonomous Driving
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-01T20%3A16%3A45IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-kuleuven&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=CLAD:%20A%20realistic%20Continual%20Learning%20benchmark%20for%20Autonomous%20Driving&rft.jtitle=NEURAL%20NETWORKS&rft.au=Verwimp,%20Eli&rft.date=2023-04&rft.volume=161&rft.spage=659&rft.epage=669&rft.pages=659-669&rft.issn=0893-6080&rft_id=info:doi/&rft_dat=%3Ckuleuven%3E20_500_12942_735519%3C/kuleuven%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true