Unsupervised Pre-Training for Detection Transformers
DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. In...
Uložené v:
| Vydané v: | IEEE transactions on pattern analysis and machine intelligence Ročník 45; číslo 11; s. 1 - 11 |
|---|---|
| Hlavní autori: | , , , |
| Médium: | Journal Article |
| Jazyk: | English |
| Vydavateľské údaje: |
New York
IEEE
01.11.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Predmet: | |
| ISSN: | 0162-8828, 1939-3539, 2160-9292, 1939-3539 |
| On-line prístup: | Získať plný text |
| Tagy: |
Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
|
| Abstract | DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose a novel pretext task named random query patch detection in Unsupervised Pre-training DETR (UP-DETR). Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the input image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade off classification and localization preferences in the pretext task, we find that freezing the CNN backbone is the prerequisite for the success of pre-training transformers. (2) To perform multi-query localization, we develop UP-DETR with multi-query patch detection with attention mask. Besides, UP-DETR also provides a unified perspective for fine-tuning object detection and one-shot detection tasks. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher average precision on object detection, one-shot detection and panoptic segmentation. Code and pre-training models: https://github.com/dddzg/up-detr . |
|---|---|
| AbstractList | DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose a novel pretext task named random query patch detection in Unsupervised Pre-training DETR (UP-DETR). Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the input image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade off classification and localization preferences in the pretext task, we find that freezing the CNN backbone is the prerequisite for the success of pre-training transformers. (2) To perform multi-query localization, we develop UP-DETR with multi-query patch detection with attention mask. Besides, UP-DETR also provides a unified perspective for fine-tuning object detection and one-shot detection tasks. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher average precision on object detection, one-shot detection and panoptic segmentation. Code and pre-training models: https://github.com/dddzg/up-detr . DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose a novel pretext task named random query patch detection in Unsupervised Pre-training DETR (UP-DETR). Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the input image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade off classification and localization preferences in the pretext task, we find that freezing the CNN backbone is the prerequisite for the success of pre-training transformers. (2) To perform multi-query localization, we develop UP-DETR with multi-query patch detection with attention mask. Besides, UP-DETR also provides a unified perspective for fine-tuning object detection and one-shot detection tasks. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher average precision on object detection, one-shot detection and panoptic segmentation. Code and pre-training models: https://github.com/dddzg/up-detr.DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose a novel pretext task named random query patch detection in Unsupervised Pre-training DETR (UP-DETR). Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the input image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade off classification and localization preferences in the pretext task, we find that freezing the CNN backbone is the prerequisite for the success of pre-training transformers. (2) To perform multi-query localization, we develop UP-DETR with multi-query patch detection with attention mask. Besides, UP-DETR also provides a unified perspective for fine-tuning object detection and one-shot detection tasks. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher average precision on object detection, one-shot detection and panoptic segmentation. Code and pre-training models: https://github.com/dddzg/up-detr. |
| Author | Chen, Junying Cai, Bolun Dai, Zhigang Lin, Yugeng |
| Author_xml | – sequence: 1 givenname: Zhigang surname: Dai fullname: Dai, Zhigang organization: School of Software Engineering, South China University of Technology, China – sequence: 2 givenname: Bolun surname: Cai fullname: Cai, Bolun organization: Tencent Wechat AI, Guangzhou, China – sequence: 3 givenname: Yugeng surname: Lin fullname: Lin, Yugeng organization: Tencent Wechat AI, Guangzhou, China – sequence: 4 givenname: Junying orcidid: 0000-0002-5614-9731 surname: Chen fullname: Chen, Junying organization: School of Software Engineering, South China University of Technology, China |
| BookMark | eNp9kE9PAjEQxRuDiYB-Ab2QePGy2E673e6R4D8SjBzg3HTL1CyBLra7Jn57ixAPHjzN5OX9JvPegPR845GQa0bHjNHyfrmYvM7GQAHGHJjMmTgj_bTQrIQSeqRPmYRMKVAXZBDjhlImcsr7RKx87PYYPuuI69EiYLYMpva1fx-5JowesEXb1o0fJdnHJO0wxEty7sw24tVpDsnq6XE5fcnmb8-z6WSeWQ6qzQTPkTpWOmWNM7biRW5kIQU3VtmCSVvkFeMV5dSWQjmBrpJJwZLKdeFYxYfk7nh3H5qPDmOrd3W0uN0aj00XNRSQzuUC8mS9_WPdNF3w6TsNqgAhGAiZXOrosqGJMaDTtm7NIV-bUm81o_pQp_6pUx_q1Kc6Ewp_0H2odyZ8_Q_dHKEaEX-BsgQJlPFvHEiByg |
| CODEN | ITPIDJ |
| CitedBy_id | crossref_primary_10_7717_peerj_cs_2107 crossref_primary_10_1016_j_media_2025_103466 crossref_primary_10_1007_s11263_024_02336_9 crossref_primary_10_1007_s10489_023_04600_w crossref_primary_10_1109_TIM_2024_3369157 crossref_primary_10_1016_j_cosrev_2024_100648 crossref_primary_10_3389_fnbot_2024_1484088 crossref_primary_10_1109_TPAMI_2024_3409826 crossref_primary_10_1038_s41598_024_80675_w crossref_primary_10_1109_JOE_2025_3529121 |
| Cites_doi | 10.1109/CVPR42600.2020.01158 10.1109/CVPR52688.2022.01420 10.1109/CVPR.2017.106 10.1109/CVPR.2019.00637 10.1109/CVPR42600.2020.00975 10.1109/CVPR42600.2020.00978 10.1109/ICCV48922.2021.00986 10.1109/CVPR.2016.90 10.1109/CVPR.2019.00754 10.1109/ICCV.2017.324 10.1109/CVPR.2016.255 10.1109/CVPR42600.2020.01020 10.1109/CVPR52688.2022.01553 10.1109/CVPR.2017.472 10.1109/CVPR.2009.5206848 10.1109/CVPR.2006.100 10.1109/ICCV48922.2021.00060 10.1109/CVPR.2018.00644 10.1109/ICCV.2019.00051 10.1007/978-3-030-58452-8_13 10.1109/CVPR.2018.00393 10.1109/CVPR.2019.00902 10.1007/s11263-009-0275-4 10.1109/ICCV.2019.00972 10.1109/ICCV.2019.00502 10.1109/CVPR.2018.00935 10.1109/CVPR.2019.00963 10.1109/ICCV.2017.322 |
| ContentType | Journal Article |
| Copyright | Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023 |
| Copyright_xml | – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023 |
| DBID | 97E RIA RIE AAYXX CITATION 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8 |
| DOI | 10.1109/TPAMI.2022.3216514 |
| DatabaseName | IEEE All-Society Periodicals Package (ASPP) 2005-present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef Computer and Information Systems Abstracts Electronics & Communications Abstracts Technology Research Database ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitle | CrossRef Technology Research Database Computer and Information Systems Abstracts – Academic Electronics & Communications Abstracts ProQuest Computer Science Collection Computer and Information Systems Abstracts Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitleList | Technology Research Database MEDLINE - Academic |
| Database_xml | – sequence: 1 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher – sequence: 2 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Engineering Computer Science |
| EISSN | 2160-9292 1939-3539 |
| EndPage | 11 |
| ExternalDocumentID | 10_1109_TPAMI_2022_3216514 9926201 |
| Genre | orig-research |
| GroupedDBID | --- -DZ -~X .DC 0R~ 29I 4.4 53G 5GY 5VS 6IK 97E 9M8 AAJGR AASAJ AAWTH ABFSI ABQJQ ABVLG ACGFO ACGFS ACIWK ACNCT ADRHT AENEX AETEA AETIX AGQYO AGSQL AHBIQ AI. AIBXA AKJIK AKQYR ALLEH ALMA_UNASSIGNED_HOLDINGS ASUFR ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ CS3 DU5 E.L EBS EJD F5P FA8 HZ~ H~9 IBMZZ ICLAB IEDLZ IFIPE IFJZH IPLJI JAVBF LAI M43 MS~ O9- OCL P2P PQQKQ RIA RIE RNI RNS RXW RZB TAE TN5 UHB VH1 ~02 AAYXX CITATION 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8 ABAZT |
| ID | FETCH-LOGICAL-c328t-435e0f19f8cafacb375a67643ac8c716c75b13b030c948f4efb65b1e906d7f1b3 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 19 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001258161200001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 0162-8828 1939-3539 |
| IngestDate | Sat Sep 27 22:17:03 EDT 2025 Sun Nov 30 04:51:58 EST 2025 Sat Nov 29 02:58:21 EST 2025 Tue Nov 18 21:58:06 EST 2025 Tue Nov 25 14:44:24 EST 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 11 |
| Language | English |
| License | https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037 |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c328t-435e0f19f8cafacb375a67643ac8c716c75b13b030c948f4efb65b1e906d7f1b3 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23 |
| ORCID | 0000-0002-5614-9731 |
| PQID | 2872441246 |
| PQPubID | 85458 |
| PageCount | 11 |
| ParticipantIDs | crossref_citationtrail_10_1109_TPAMI_2022_3216514 crossref_primary_10_1109_TPAMI_2022_3216514 ieee_primary_9926201 proquest_miscellaneous_2727645425 proquest_journals_2872441246 |
| PublicationCentury | 2000 |
| PublicationDate | 2023-11-01 |
| PublicationDateYYYYMMDD | 2023-11-01 |
| PublicationDate_xml | – month: 11 year: 2023 text: 2023-11-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationPlace | New York |
| PublicationPlace_xml | – name: New York |
| PublicationTitle | IEEE transactions on pattern analysis and machine intelligence |
| PublicationTitleAbbrev | TPAMI |
| PublicationYear | 2023 |
| Publisher | IEEE The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Publisher_xml | – name: IEEE – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| References | ref13 ref53 ref52 ref10 bertinetto (ref47) 2016 ref19 loshchilov (ref40) 2018 liu (ref21) 2016 li (ref51) 2020 caron (ref17) 2018 ref46 ref45 ref48 ref42 ref41 ref44 ref43 ren (ref4) 2015 asano (ref18) 2019 chen (ref11) 2020 ref3 ref6 yang (ref29) 2019 zhang (ref49) 2019 carion (ref1) 2020 grill (ref15) 2020 lin (ref5) 2014 ref34 ref37 ref36 chen (ref14) 2020 ref31 ref30 hsieh (ref50) 2019 ref32 radford (ref8) 2019; 1 radford (ref7) 0 ref39 han (ref35) 2021 vaswani (ref2) 2017 zhou (ref23) 2019 ref26 cao (ref16) 0 ref25 ref20 devlin (ref9) 2018 ref22 dosovitskiy (ref33) 2021 caron (ref12) 2020 law (ref24) 2018 ref28 ref27 dong (ref38) 2019 |
| References_xml | – start-page: 21271 year: 2020 ident: ref15 article-title: Bootstrap your own latent-a new approach to self-supervised learning publication-title: Proc Adv Neural Inf Process Syst – ident: ref28 doi: 10.1109/CVPR42600.2020.01158 – start-page: 9912 year: 2020 ident: ref12 article-title: Unsupervised learning of visual features by contrasting cluster assignments publication-title: Proc Adv Neural Inf Process Syst – ident: ref41 doi: 10.1109/CVPR52688.2022.01420 – start-page: 5998 year: 2017 ident: ref2 article-title: Attention is all you need publication-title: Proc Adv Neural Inf Process Syst – ident: ref43 doi: 10.1109/CVPR.2017.106 – year: 2019 ident: ref49 article-title: Comparison network for one-shot conditional object detection – ident: ref30 doi: 10.1109/CVPR.2019.00637 – ident: ref10 doi: 10.1109/CVPR42600.2020.00975 – year: 2019 ident: ref50 article-title: One-shot object detection with co-attention and co-excitation – year: 2020 ident: ref11 article-title: Improved baselines with momentum contrastive learning – start-page: 850 year: 2016 ident: ref47 article-title: Fully-convolutional siamese networks for object tracking publication-title: Proc Eur Conf Comput Vis – ident: ref3 doi: 10.1109/CVPR42600.2020.00978 – ident: ref36 doi: 10.1109/ICCV48922.2021.00986 – start-page: 132 year: 2018 ident: ref17 article-title: Deep clustering for unsupervised learning of visual features publication-title: Proc Eur Conf Comput Vis – ident: ref13 doi: 10.1109/CVPR.2016.90 – ident: ref45 doi: 10.1109/CVPR.2019.00754 – ident: ref20 doi: 10.1109/ICCV.2017.324 – start-page: 1 year: 2018 ident: ref40 article-title: Decoupled weight decay regularization publication-title: Proc Int Conf Learn Representations – ident: ref25 doi: 10.1109/CVPR.2016.255 – ident: ref27 doi: 10.1109/CVPR42600.2020.01020 – volume: 1 year: 2019 ident: ref8 article-title: Language models are unsupervised multitask learners publication-title: OpenAIRE blog – ident: ref37 doi: 10.1109/CVPR52688.2022.01553 – year: 0 ident: ref7 article-title: Improving language understanding by generative pre-training – ident: ref42 doi: 10.1109/CVPR.2017.472 – year: 2020 ident: ref14 article-title: A simple framework for contrastive learning of visual representations – start-page: 13 063 year: 2019 ident: ref38 article-title: Unified language model pre-training for natural language understanding and generation publication-title: Proc Adv Neural Inf Process Syst – ident: ref39 doi: 10.1109/CVPR.2009.5206848 – ident: ref32 doi: 10.1109/CVPR.2006.100 – start-page: 91 year: 2015 ident: ref4 article-title: Faster R-CNN: Towards real-time object detection with region proposal networks publication-title: Proc Adv Neural Inf Process Syst – start-page: 740 year: 2014 ident: ref5 article-title: Microsoft COCO: Common objects in context publication-title: Proc Eur Conf Comput Vis – ident: ref34 doi: 10.1109/ICCV48922.2021.00060 – ident: ref19 doi: 10.1109/CVPR.2018.00644 – ident: ref26 doi: 10.1109/ICCV.2019.00051 – year: 2020 ident: ref1 article-title: End-to-end object detection with transformers doi: 10.1007/978-3-030-58452-8_13 – ident: ref31 doi: 10.1109/CVPR.2018.00393 – year: 2019 ident: ref23 article-title: Objects as points – start-page: 5753 year: 2019 ident: ref29 article-title: XLNet: Generalized autoregressive pretraining for language understanding publication-title: Proc Adv Neural Inf Process Syst – ident: ref53 doi: 10.1109/CVPR.2019.00902 – year: 2018 ident: ref9 article-title: Bert: Pre-training of deep bidirectional transformers for language understanding – ident: ref6 doi: 10.1007/s11263-009-0275-4 – ident: ref22 doi: 10.1109/ICCV.2019.00972 – start-page: 1 year: 2021 ident: ref33 article-title: An image is worth 16x16 words: Transformers for image recognition at scale publication-title: Proc Int Conf Learn Representations – ident: ref46 doi: 10.1109/ICCV.2019.00502 – start-page: 15908 year: 2021 ident: ref35 article-title: Transformer in transformer publication-title: Proc Adv Neural Inf Process Syst – ident: ref48 doi: 10.1109/CVPR.2018.00935 – start-page: 734 year: 2018 ident: ref24 article-title: CornerNet: Detecting objects as paired keypoints publication-title: Proc Eur Conf Comput Vis – start-page: 1 year: 2019 ident: ref18 article-title: Self-labelling via simultaneous clustering and representation learning publication-title: Proc Int Conf Learn Representations – start-page: 21 year: 2016 ident: ref21 article-title: SSD: Single shot multibox detector publication-title: Proc Eur Conf Comput Vis – start-page: 15614 year: 0 ident: ref16 article-title: Parametric instance classification for unsupervised visual feature learning publication-title: Proc Adv Neural Inf Process Syst – year: 2020 ident: ref51 article-title: One-shot object detection without fine-tuning – ident: ref52 doi: 10.1109/CVPR.2019.00963 – ident: ref44 doi: 10.1109/ICCV.2017.322 |
| SSID | ssj0014503 |
| Score | 2.5500627 |
| Snippet | DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture.... |
| SourceID | proquest crossref ieee |
| SourceType | Aggregation Database Enrichment Source Index Database Publisher |
| StartPage | 1 |
| SubjectTerms | Coders Decoding Encoders-Decoders Feature extraction Freezing Image segmentation Localization Location awareness Natural language processing Object detection Object recognition one-shot detection Queries self-supervised learning Task analysis Training transformer Transformers unsupervised pre-training |
| Title | Unsupervised Pre-Training for Detection Transformers |
| URI | https://ieeexplore.ieee.org/document/9926201 https://www.proquest.com/docview/2872441246 https://www.proquest.com/docview/2727645425 |
| Volume | 45 |
| WOSCitedRecordID | wos001258161200001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 2160-9292 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0014503 issn: 0162-8828 databaseCode: RIE dateStart: 19790101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8QwEB5UPOjBt7i-qOBNq02TNs1RfKAHZQ8r7K2kyQQE6S778Pc7ybZFUQRvpZ2WMslkvslk5gM4t4ZQuKhkzLQysXDIYo1SxEapyjnygFbbQDYhX16K4VD1l-Cyq4VBxHD4DK_8Zcjl25GZ-62ya-Wb2_lirWUp80WtVpcxEFlgQSYEQxZOYURbIJOo60H_5vmJQsE0veIpyzMmvjmhwKryYykO_uVh839_tgUbDY6MbhYDvw1LWO_AZsvREDUmuwPrXxoO7oJ4rafzsV8epmij_gTjQUMRERF4je5wFk5m1dGgxbOEDvfg9eF-cPsYN7wJseFpMYsJAWHimHKF0U6bistM55KghzaFofjIyKxivCLzNkoUTqCrcrqDKsmtdKzi-7BSj2o8gIg5yRXjmRHKCmdzipe5tRnXjiMqnvSAtYosTdNU3HNbvJchuEhUGZRfeuWXjfJ7cNG9M1601PhTeteru5NsNN2D43a8ysbqpiVFf4RWCLHkPTjrHpO9-CSIrnE0JxkCbL6LWZod_v7lI1jzlPKLesNjWJlN5ngCq-Zj9jadnNLUGxanYep9AgHr1LQ |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1bS-UwEB5EhV0f1PWCx2sF39Zq06SXPIoXFPVwHir4VtJkAgtLj5yLv99JTlpcdhH2rbTTUiad5Ps6mfkAzowmFC6aImZK6lhYZLHCQsRaysZaWgGNMl5sohgOy9dXOVqC874WBhH95jO8cIc-l2_Geu5-lV1K19zOFWutOOWsUK3V5wxE5nWQCcNQjBOR6EpkEnlZja6eH4gMpukFT1meMfHHMuR1Vf6ajP0Kc7fxf--2CesBSUZXi6H_AUvYbsFGp9IQhaDdgrVPLQe3Qby00_mbmyCmaKLRBOMqiEREBF-jG5z5vVltVHWIlvDhDrzc3VbX93FQTog1T8tZTBgIE8ukLbWySje8yFReEPhQutTEkHSRNYw3FOBaitIKtE1OZ1AmuSksa_guLLfjFvcgYrbgkvFMC2mENTkxZm5MxpXliJInA2CdI2sd2oo7dYvftacXiay982vn_Do4fwA_-3veFk01vrTedu7uLYOnB3DYjVcd4m5aE_8jvEKYJR_AaX-ZIsalQVSL4znZEGRzfczSbP_fTz6Bb_fV81P99DB8PIDvTmB-UX14CMuzyRyPYFW_z35NJ8f-A_wAPODXFQ |
| 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=Unsupervised+Pre-Training+for+Detection+Transformers&rft.jtitle=IEEE+transactions+on+pattern+analysis+and+machine+intelligence&rft.au=Dai%2C+Zhigang&rft.au=Cai%2C+Bolun&rft.au=Lin%2C+Yugeng&rft.au=Chen%2C+Junying&rft.date=2023-11-01&rft.issn=0162-8828&rft.eissn=2160-9292&rft.spage=1&rft.epage=11&rft_id=info:doi/10.1109%2FTPAMI.2022.3216514&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_TPAMI_2022_3216514 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0162-8828&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0162-8828&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0162-8828&client=summon |