Cross Domain Object Detection by Target-Perceived Dual Branch Distillation

Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even thou...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: He, Mengzhe, Wang, Yali, Wu, Jiaxi, Wang, Yiru, Li, Hanqing, Li, Bo, Gan, Weihao, Wu, Wei, Qiao, Yu
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator He, Mengzhe
Wang, Yali
Wu, Jiaxi
Wang, Yiru
Li, Hanqing
Li, Bo
Gan, Weihao
Wu, Wei
Qiao, Yu
description Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even though they are closely coupled in cross domain object detection. To solve this problem, we propose a novel Target-perceived Dual-branch Distillation (TDD) framework. By integrating detection branches of both source and target domains in a unified teacher-student learning scheme, it can reduce domain shift and generate reliable supervision effectively. In particular, we first introduce a distinct Target Proposal Perceiver between two domains. It can adaptively enhance source detector to perceive objects in a target image, by leveraging target proposal contexts from iterative cross-attention. Afterwards, we design a concise Dual Branch Self Distillation strategy for model training, which can progressively integrate complementary object knowledge from different domains via self-distillation in two branches. Finally, we conduct extensive experiments on a number of widely-used scenarios in cross domain object detection. The results show that our TDD significantly outperforms the state-of-the-art methods on all the benchmarks. Our code and model will be available at https://github.com/Feobi1999/TDD.
doi_str_mv 10.48550/arxiv.2205.01291
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2205_01291</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2205_01291</sourcerecordid><originalsourceid>FETCH-LOGICAL-a671-322d103c882da7818c9f9950556c7b2c44a80b1ada46dd6524ebb50557f996ae3</originalsourceid><addsrcrecordid>eNotj7FuwjAURb10qCgf0Kn-gaS2YzvO2CZAqZDokD16th-tq5BUTkDw9yXAdIZ7dKVDyDNnqTRKsVeIp3BMhWAqZVwU_JF8lrEfBlr1ewgd3dpfdCOtcLwg9B21Z1pD_MYx-cLoMBzR0-oALX2P0LkfWoVhDG0Lk_xEHnbQDji_c0bq5aIuP5LNdrUu3zYJ6JwnmRCes8wZIzzkhhtX7IpCMaW0y61wUoJhloMHqb3XSki0dprzi6YBsxl5ud1eY5q_GPYQz80U1Vyjsn_BiEaF</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Cross Domain Object Detection by Target-Perceived Dual Branch Distillation</title><source>arXiv.org</source><creator>He, Mengzhe ; Wang, Yali ; Wu, Jiaxi ; Wang, Yiru ; Li, Hanqing ; Li, Bo ; Gan, Weihao ; Wu, Wei ; Qiao, Yu</creator><creatorcontrib>He, Mengzhe ; Wang, Yali ; Wu, Jiaxi ; Wang, Yiru ; Li, Hanqing ; Li, Bo ; Gan, Weihao ; Wu, Wei ; Qiao, Yu</creatorcontrib><description>Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even though they are closely coupled in cross domain object detection. To solve this problem, we propose a novel Target-perceived Dual-branch Distillation (TDD) framework. By integrating detection branches of both source and target domains in a unified teacher-student learning scheme, it can reduce domain shift and generate reliable supervision effectively. In particular, we first introduce a distinct Target Proposal Perceiver between two domains. It can adaptively enhance source detector to perceive objects in a target image, by leveraging target proposal contexts from iterative cross-attention. Afterwards, we design a concise Dual Branch Self Distillation strategy for model training, which can progressively integrate complementary object knowledge from different domains via self-distillation in two branches. Finally, we conduct extensive experiments on a number of widely-used scenarios in cross domain object detection. The results show that our TDD significantly outperforms the state-of-the-art methods on all the benchmarks. Our code and model will be available at https://github.com/Feobi1999/TDD.</description><identifier>DOI: 10.48550/arxiv.2205.01291</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2022-05</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2205.01291$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2205.01291$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>He, Mengzhe</creatorcontrib><creatorcontrib>Wang, Yali</creatorcontrib><creatorcontrib>Wu, Jiaxi</creatorcontrib><creatorcontrib>Wang, Yiru</creatorcontrib><creatorcontrib>Li, Hanqing</creatorcontrib><creatorcontrib>Li, Bo</creatorcontrib><creatorcontrib>Gan, Weihao</creatorcontrib><creatorcontrib>Wu, Wei</creatorcontrib><creatorcontrib>Qiao, Yu</creatorcontrib><title>Cross Domain Object Detection by Target-Perceived Dual Branch Distillation</title><description>Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even though they are closely coupled in cross domain object detection. To solve this problem, we propose a novel Target-perceived Dual-branch Distillation (TDD) framework. By integrating detection branches of both source and target domains in a unified teacher-student learning scheme, it can reduce domain shift and generate reliable supervision effectively. In particular, we first introduce a distinct Target Proposal Perceiver between two domains. It can adaptively enhance source detector to perceive objects in a target image, by leveraging target proposal contexts from iterative cross-attention. Afterwards, we design a concise Dual Branch Self Distillation strategy for model training, which can progressively integrate complementary object knowledge from different domains via self-distillation in two branches. Finally, we conduct extensive experiments on a number of widely-used scenarios in cross domain object detection. The results show that our TDD significantly outperforms the state-of-the-art methods on all the benchmarks. Our code and model will be available at https://github.com/Feobi1999/TDD.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj7FuwjAURb10qCgf0Kn-gaS2YzvO2CZAqZDokD16th-tq5BUTkDw9yXAdIZ7dKVDyDNnqTRKsVeIp3BMhWAqZVwU_JF8lrEfBlr1ewgd3dpfdCOtcLwg9B21Z1pD_MYx-cLoMBzR0-oALX2P0LkfWoVhDG0Lk_xEHnbQDji_c0bq5aIuP5LNdrUu3zYJ6JwnmRCes8wZIzzkhhtX7IpCMaW0y61wUoJhloMHqb3XSki0dprzi6YBsxl5ud1eY5q_GPYQz80U1Vyjsn_BiEaF</recordid><startdate>20220502</startdate><enddate>20220502</enddate><creator>He, Mengzhe</creator><creator>Wang, Yali</creator><creator>Wu, Jiaxi</creator><creator>Wang, Yiru</creator><creator>Li, Hanqing</creator><creator>Li, Bo</creator><creator>Gan, Weihao</creator><creator>Wu, Wei</creator><creator>Qiao, Yu</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20220502</creationdate><title>Cross Domain Object Detection by Target-Perceived Dual Branch Distillation</title><author>He, Mengzhe ; Wang, Yali ; Wu, Jiaxi ; Wang, Yiru ; Li, Hanqing ; Li, Bo ; Gan, Weihao ; Wu, Wei ; Qiao, Yu</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a671-322d103c882da7818c9f9950556c7b2c44a80b1ada46dd6524ebb50557f996ae3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>He, Mengzhe</creatorcontrib><creatorcontrib>Wang, Yali</creatorcontrib><creatorcontrib>Wu, Jiaxi</creatorcontrib><creatorcontrib>Wang, Yiru</creatorcontrib><creatorcontrib>Li, Hanqing</creatorcontrib><creatorcontrib>Li, Bo</creatorcontrib><creatorcontrib>Gan, Weihao</creatorcontrib><creatorcontrib>Wu, Wei</creatorcontrib><creatorcontrib>Qiao, Yu</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>He, Mengzhe</au><au>Wang, Yali</au><au>Wu, Jiaxi</au><au>Wang, Yiru</au><au>Li, Hanqing</au><au>Li, Bo</au><au>Gan, Weihao</au><au>Wu, Wei</au><au>Qiao, Yu</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Cross Domain Object Detection by Target-Perceived Dual Branch Distillation</atitle><date>2022-05-02</date><risdate>2022</risdate><abstract>Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even though they are closely coupled in cross domain object detection. To solve this problem, we propose a novel Target-perceived Dual-branch Distillation (TDD) framework. By integrating detection branches of both source and target domains in a unified teacher-student learning scheme, it can reduce domain shift and generate reliable supervision effectively. In particular, we first introduce a distinct Target Proposal Perceiver between two domains. It can adaptively enhance source detector to perceive objects in a target image, by leveraging target proposal contexts from iterative cross-attention. Afterwards, we design a concise Dual Branch Self Distillation strategy for model training, which can progressively integrate complementary object knowledge from different domains via self-distillation in two branches. Finally, we conduct extensive experiments on a number of widely-used scenarios in cross domain object detection. The results show that our TDD significantly outperforms the state-of-the-art methods on all the benchmarks. Our code and model will be available at https://github.com/Feobi1999/TDD.</abstract><doi>10.48550/arxiv.2205.01291</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2205.01291
ispartof
issn
language eng
recordid cdi_arxiv_primary_2205_01291
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
title Cross Domain Object Detection by Target-Perceived Dual Branch Distillation
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-27T05%3A45%3A18IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Cross%20Domain%20Object%20Detection%20by%20Target-Perceived%20Dual%20Branch%20Distillation&rft.au=He,%20Mengzhe&rft.date=2022-05-02&rft_id=info:doi/10.48550/arxiv.2205.01291&rft_dat=%3Carxiv_GOX%3E2205_01291%3C/arxiv_GOX%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