在线推断校准的小样本目标检测

TP391.4; 针对少量样本条件下模型易过拟合、目标错检与漏检问题,本文基于TFA(two-stage fine-tuning approach)提出了一种在线推断校准的小样本目标检测框架.该框架设计了一种全新的Attention-FPN网络,通过建模特征通道间的依赖关系选择性融合特征,结合分级冻结的学习机制引导RPN模块提取正确的新类前景目标;同时,构建了一种在线校准模块对样本进行实例分割编码,对众多候选目标进行评分重加权处理,纠正误检和漏检的预测目标.结果表明,所提算法在VOC数据集Novel Set1中,五个任务的平均nAP50提升10.16％,在性能上优于目前的主流算法....

Full description

Saved in:
Bibliographic Details
Published in光电工程 Vol. 50; no. 1; pp. 83 - 97
Main Authors 彭昊, 王婉祺, 陈龙, 彭先蓉, 张建林, 徐智勇, 魏宇星, 李美惠
Format Journal Article
LanguageChinese
Published 中国科学院大学,北京 100049%中国科学院光电技术研究所,四川成都 610209 2023
中国科学院光电技术研究所,四川成都 610209
Subjects
小样本目标检测
特征通道
RPN
在线校准
分级冻结
Attention-FPN
Online AccessGet full text
ISSN1003-501X
DOI10.12086/oee.2023.220180

Cover

Abstract TP391.4; 针对少量样本条件下模型易过拟合、目标错检与漏检问题,本文基于TFA(two-stage fine-tuning approach)提出了一种在线推断校准的小样本目标检测框架.该框架设计了一种全新的Attention-FPN网络,通过建模特征通道间的依赖关系选择性融合特征,结合分级冻结的学习机制引导RPN模块提取正确的新类前景目标;同时,构建了一种在线校准模块对样本进行实例分割编码,对众多候选目标进行评分重加权处理,纠正误检和漏检的预测目标.结果表明,所提算法在VOC数据集Novel Set1中,五个任务的平均nAP50提升10.16％,在性能上优于目前的主流算法.
AbstractList TP391.4; 针对少量样本条件下模型易过拟合、目标错检与漏检问题,本文基于TFA(two-stage fine-tuning approach)提出了一种在线推断校准的小样本目标检测框架.该框架设计了一种全新的Attention-FPN网络,通过建模特征通道间的依赖关系选择性融合特征,结合分级冻结的学习机制引导RPN模块提取正确的新类前景目标;同时,构建了一种在线校准模块对样本进行实例分割编码,对众多候选目标进行评分重加权处理,纠正误检和漏检的预测目标.结果表明,所提算法在VOC数据集Novel Set1中,五个任务的平均nAP50提升10.16％,在性能上优于目前的主流算法.
Author 王婉祺
陈龙
李美惠
魏宇星
彭昊
张建林
彭先蓉
徐智勇
AuthorAffiliation 中国科学院光电技术研究所,四川成都 610209;中国科学院大学,北京 100049%中国科学院光电技术研究所,四川成都 610209
AuthorAffiliation_xml – name: 中国科学院光电技术研究所,四川成都 610209;中国科学院大学,北京 100049%中国科学院光电技术研究所,四川成都 610209
Author_FL Peng Xianrong
Xu Zhiyong
Zhang Jianlin
Wang Wanqi
Peng Hao
Li Meihui
Chen Long
Wei Yuxing
Author_FL_xml – sequence: 1
  fullname: Peng Hao
– sequence: 2
  fullname: Wang Wanqi
– sequence: 3
  fullname: Chen Long
– sequence: 4
  fullname: Peng Xianrong
– sequence: 5
  fullname: Zhang Jianlin
– sequence: 6
  fullname: Xu Zhiyong
– sequence: 7
  fullname: Wei Yuxing
– sequence: 8
  fullname: Li Meihui
Author_xml – sequence: 1
  fullname: 彭昊
– sequence: 2
  fullname: 王婉祺
– sequence: 3
  fullname: 陈龙
– sequence: 4
  fullname: 彭先蓉
– sequence: 5
  fullname: 张建林
– sequence: 6
  fullname: 徐智勇
– sequence: 7
  fullname: 魏宇星
– sequence: 8
  fullname: 李美惠
BookMark eNrjYmDJy89LZWCQMDTQMzQysDDTz09N1TMyMDLWMzIyMLQwYGHgNDQwMNY1NTCM4GDgLS7OTDIwMDGyNLcwM-dk0Ho6Z8XzXfuf9a14Nm3tswULn7a3PZ_V8nRD_7MF25_NWfN89rpnC9qfLW54trWbh4E1LTGnOJUXSnMzhLq5hjh76Pr4u3s6O_roFgPtMddNNEwyTLE0TjUxs0hLTDJLTjOySDRLS7ZMMjcxtDC1NEhOTjRONLcwMQESyWmmFmlmSYnmiSnGSZamlqmJqWbG3AyqEHPLE_PSEvPS47PyS4vygDbGp6ekJ4N8ZgCyyBgAfElToA
ClassificationCodes TP391.4
ContentType Journal Article
Copyright Copyright © Wanfang Data Co. Ltd. All Rights Reserved.
Copyright_xml – notice: Copyright © Wanfang Data Co. Ltd. All Rights Reserved.
DBID 2B.
4A8
92I
93N
PSX
TCJ
DOI 10.12086/oee.2023.220180
DatabaseName Wanfang Data Journals - Hong Kong
WANFANG Data Centre
Wanfang Data Journals
万方数据期刊 - 香港版
China Online Journals (COJ)
China Online Journals (COJ)
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
DocumentTitle_FL Few-shot object detection via online inferential calibration
EndPage 97
ExternalDocumentID gdgc202301007
GrantInformation_xml – fundername: 国家自然科学基金
  funderid: (62101529)
GroupedDBID 2B.
4A8
8FE
8FG
92I
93N
ABJCF
ABUWG
AFKRA
ALMA_UNASSIGNED_HOLDINGS
BENPR
BGLVJ
BPHCQ
BVBZV
CCPQU
HCIFZ
L6V
M7S
PHGZM
PHGZT
PIMPY
PMFND
PQQKQ
PROAC
PSX
PTHSS
TCJ
ID FETCH-LOGICAL-s1007-a1b1d93e468fab6cf28a6fc9b7418590cca3a7844a78cf58f6ba7ad3b959eae63
ISSN 1003-501X
IngestDate Thu May 29 03:55:49 EDT 2025
IsPeerReviewed true
IsScholarly true
Issue 1
Keywords 小样本目标检测
特征通道
RPN
在线校准
分级冻结
Attention-FPN
Language Chinese
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-s1007-a1b1d93e468fab6cf28a6fc9b7418590cca3a7844a78cf58f6ba7ad3b959eae63
PageCount 15
ParticipantIDs wanfang_journals_gdgc202301007
PublicationCentury 2000
PublicationDate 2023
PublicationDateYYYYMMDD 2023-01-01
PublicationDate_xml – year: 2023
  text: 2023
PublicationDecade 2020
PublicationTitle 光电工程
PublicationTitle_FL Opto-Electronic Engineering
PublicationYear 2023
Publisher 中国科学院大学,北京 100049%中国科学院光电技术研究所,四川成都 610209
中国科学院光电技术研究所,四川成都 610209
Publisher_xml – name: 中国科学院光电技术研究所,四川成都 610209
– name: 中国科学院大学,北京 100049%中国科学院光电技术研究所,四川成都 610209
SSID ssib004297867
ssib023646518
ssib036437391
ssib023167165
ssj0002964646
ssib002258422
ssib001102639
ssib051369860
ssib000459782
Score 2.3742266
Snippet TP391.4; 针对少量样本条件下模型易过拟合、目标错检与漏检问题,本文基于TFA(two-stage fine-tuning approach)提出了一种在线推断校准的小样本目标检测框架.该框架设计了一...
SourceID wanfang
SourceType Aggregation Database
StartPage 83
Title 在线推断校准的小样本目标检测
URI https://d.wanfangdata.com.cn/periodical/gdgc202301007
Volume 50
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVPQU
  databaseName: ProQuest Technology Collection
  issn: 1003-501X
  databaseCode: 8FG
  dateStart: 20180101
  customDbUrl:
  isFulltext: true
  dateEnd: 99991231
  titleUrlDefault: https://search.proquest.com/technologycollection1
  omitProxy: true
  ssIdentifier: ssj0002964646
  providerName: ProQuest
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnR09b9UwMCplYUEgQHxWHfBClZLYieMbndc8KiSYWqlb5eQlj-lVoq9LJwZEByTYEIIOD7F0AMSCkFj4NS_vb3DnmJe0VIIiPUXW3fl8vnuxz_H57Hl3B2FRDcrY4NpEST8qDPfBBKUvwOD8Iqo8tju6jx7L9c3o4Va8tXDubSdqaW-crxb7p54r-R-rIgztSqdkz2DZOVMEYBnti0-0MD7_ycYsixn0mFYsS1iqWdpnmWQqsxDJQDK9RgUdMB0SMa78lSRi0ExFBEkDpvqOJk1sLWTYszQp05lDKYvSggIjsJAiq7Tr11rm-ANbMSICYp4wHRME5VHzvR2LWXOiAWJ0i0lIepUSiQbHDlmkHRJggBhFhTRjAKewbWRBEsUARYbup43m3LH9G7IMBVWuBvYWa1Nz2AGrLIRr6drT3EIiws5RvGeb6jFILCuNilsJ7XKI8fjM3E_qDy2pnb7JJn1bPSAsFbD30tIA0TSiYCuQOsWDVS8qAQJqRQkUYAW9WB5AZyKimME4sNcJzWeqJkXvsTeymXaau4CcA9PEO_8xNfLAbtXtlJQclotVzil3W-sGzIMzh4NhQRRBaNM0nOf0pYtyMPQfdBcC0L0HAZ1ILjv5ZXGWUFEnMSR6PYlqHV1O-RfCdn-dbjGQceslCdpOFm0WJxw1JCi3kCcfi4IFZHN08LeiXAwCdfP-iU7aA3mjyoyGHd9x45J30S36lnXzBl_2FvafXPHuTQ-PZj9-1q-O6jef68mH6cGL2bvn06-v68n3-vDT7P2XenJQf3xWf3t51dvsZxu9dd_dXOLvktJ8E-bhAEQZSVWZXBYVV0ZWBeQ2VxQEOGwKk6gowkdRxaqSuUnMQOQQQ2lKKa55i6OdUXndWx7kcRIaXkVxUkRoElWiW1AZkeRVwWUub3hLrm_bbmTa3T5mwZt_I7jlXaBy813xtrc4frpX3kFPe5wvWaP_AtXHkLM
linkProvider ProQuest
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=%E5%9C%A8%E7%BA%BF%E6%8E%A8%E6%96%AD%E6%A0%A1%E5%87%86%E7%9A%84%E5%B0%8F%E6%A0%B7%E6%9C%AC%E7%9B%AE%E6%A0%87%E6%A3%80%E6%B5%8B&rft.jtitle=%E5%85%89%E7%94%B5%E5%B7%A5%E7%A8%8B&rft.au=%E5%BD%AD%E6%98%8A&rft.au=%E7%8E%8B%E5%A9%89%E7%A5%BA&rft.au=%E9%99%88%E9%BE%99&rft.au=%E5%BD%AD%E5%85%88%E8%93%89&rft.date=2023&rft.pub=%E4%B8%AD%E5%9B%BD%E7%A7%91%E5%AD%A6%E9%99%A2%E5%A4%A7%E5%AD%A6%2C%E5%8C%97%E4%BA%AC+100049%25%E4%B8%AD%E5%9B%BD%E7%A7%91%E5%AD%A6%E9%99%A2%E5%85%89%E7%94%B5%E6%8A%80%E6%9C%AF%E7%A0%94%E7%A9%B6%E6%89%80%2C%E5%9B%9B%E5%B7%9D%E6%88%90%E9%83%BD+610209&rft.issn=1003-501X&rft.volume=50&rft.issue=1&rft.spage=83&rft.epage=97&rft_id=info:doi/10.12086%2Foee.2023.220180&rft.externalDocID=gdgc202301007
thumbnail_s http://utb.summon.serialssolutions.com/2.0.0/image/custom?url=http%3A%2F%2Fwww.wanfangdata.com.cn%2Fimages%2FPeriodicalImages%2Fgdgc%2Fgdgc.jpg