MixFace: Improving face verification with a focus on fine‐grained conditions
The performance of face recognition (FR) has reached a plateau for public benchmark datasets, such as labeled faces in the wild (LFW), celebrities in frontal‐profile in the wild (CFP‐FP), and the first manually collected, in‐the‐wild age database (AgeDB), owing to the rapid advances in convolutional...
Saved in:
| Published in | ETRI journal Vol. 46; no. 4; pp. 660 - 670 |
|---|---|
| Main Authors | , , , , , |
| Format | Journal Article |
| Language | English |
| Published |
Electronics and Telecommunications Research Institute (ETRI)
01.08.2024
한국전자통신연구원 |
| Subjects | |
| Online Access | Get full text |
| ISSN | 1225-6463 2233-7326 |
| DOI | 10.4218/etrij.2023-0167 |
Cover
Loading…
| Abstract | The performance of face recognition (FR) has reached a plateau for public benchmark datasets, such as labeled faces in the wild (LFW), celebrities in frontal‐profile in the wild (CFP‐FP), and the first manually collected, in‐the‐wild age database (AgeDB), owing to the rapid advances in convolutional neural networks (CNNs). However, the effects of faces under various fine‐grained conditions on FR models have not been investigated, owing to the absence of relevant datasets. This paper analyzes their effects under different conditions and loss functions using K‐FACE, a recently introduced FR dataset with fine‐grained conditions. We propose a novel loss function called MixFace, which combines classification and metric losses. The superiority of MixFace in terms of effectiveness and robustness was experimentally demonstrated using various benchmark datasets. |
|---|---|
| AbstractList | The performance of face recognition (FR) has reached a plateau for public benchmark datasets, such as labeled faces in the wild (LFW), celebrities in frontal‐profile in the wild (CFP‐FP), and the first manually collected, in‐the‐wild age database (AgeDB), owing to the rapid advances in convolutional neural networks (CNNs). However, the effects of faces under various fine‐grained conditions on FR models have not been investigated, owing to the absence of relevant datasets. This paper analyzes their effects under different conditions and loss functions using K‐FACE, a recently introduced FR dataset with fine‐grained conditions. We propose a novel loss function called MixFace, which combines classification and metric losses. The superiority of MixFace in terms of effectiveness and robustness was experimentally demonstrated using various benchmark datasets. The performance of face recognition (FR) has reached a plateau for publicbenchmark datasets, such as labeled faces in the wild (LFW), celebrities in frontal-profile in the wild (CFP-FP), and the first manually collected, inthe-wild age database (AgeDB), owing to the rapid advances in convolutional neural networks (CNNs). However, the effects of faces under various finegrained conditions on FR models have not been investigated, owing to the absence of relevant datasets. This paper analyzes their effects under different conditions and loss functions using K-FACE, a recently introduced FR dataset with fine-grained conditions. We propose a novel loss function called MixFace, which combines classification and metric losses. The superiority of MixFace in terms of effectiveness and robustness was experimentally demonstrated using various benchmark datasets. The performance of face recognition (FR) has reached a plateau for publicbenchmark datasets, such as labeled faces in the wild (LFW), celebrities in frontal-profile in the wild (CFP-FP), and the first manually collected, inthe-wild age database (AgeDB), owing to the rapid advances in convolutional neural networks (CNNs). However, the effects of faces under various finegrained conditions on FR models have not been investigated, owing to the absence of relevant datasets. This paper analyzes their effects under different conditions and loss functions using K-FACE, a recently introduced FR dataset with fine-grained conditions. We propose a novel loss function called MixFace, which combines classification and metric losses. The superiority of MixFace in terms of effectiveness and robustness was experimentally demonstrated using various benchmark datasets. KCI Citation Count: 0 |
| Author | Park, Joochan Oh, Heung‐Seon Park, Yongjun Son, Sungbin Lee, Seonhoon Jung, Junuk |
| Author_xml | – sequence: 1 givenname: Junuk surname: Jung fullname: Jung, Junuk organization: Korea University of Technology and Education – sequence: 2 givenname: Sungbin orcidid: 0000-0001-6530-5735 surname: Son fullname: Son, Sungbin organization: Korea University of Technology and Education – sequence: 3 givenname: Joochan surname: Park fullname: Park, Joochan organization: Korea University of Technology and Education – sequence: 4 givenname: Yongjun surname: Park fullname: Park, Yongjun organization: Korea University of Technology and Education – sequence: 5 givenname: Seonhoon surname: Lee fullname: Lee, Seonhoon organization: Korea University of Technology and Education – sequence: 6 givenname: Heung‐Seon orcidid: 0000-0002-9193-8998 surname: Oh fullname: Oh, Heung‐Seon email: ohhs@koreatech.ac.kr organization: Korea University of Technology and Education |
| BackLink | https://www.kci.go.kr/kciportal/ci/sereArticleSearch/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART003106932$$DAccess content in National Research Foundation of Korea (NRF) |
| BookMark | eNqFkctOwzAQRS0EEuWxZps1UsAeP5KwQ4hHJR5SVdbWxBkXtyVGTqGw4xP4Rr6EtEVsWd2Z0b1nFnePbbexJcaOBD9RIMpTWqQwPQEOMufCFFtsACBlXkgw22wgAHRulJG7bK_rppwDV7ocsPu78H6Fjs6y4fNLim-hnWS-37M3SsEHh4sQ22wZFk8ZZj661y7rdx9a-v78miTshyZzsW3CytgdsB2P844Of3WfPV5dji9u8tuH6-HF-W3uZKlUjsCl8VqJpuLkfe2aUmMNjS8qZ8gT1LWBuvS64KglaSEVGjKcuDaiAi_32fGG2yZvZy7YiGGtk2hnyZ6PxkMruC6NVFVvHm7MTcSpfUnhGdPHOrE-xDSxmBbBzcn2TypCcFhho1yDFThCUylyEgrlyp51umG5FLsukf_jCW5XPdh1D3bVg1310CfMJrEMc_r4z24vxyMQYLSSP-MAkDA |
| Cites_doi | 10.1109/CVPR.2005.202 10.1109/CVPR.2006.100 10.1109/CVPR.2017.713 10.1109/ICCV.2017.307 10.1109/CVPRW.2017.250 10.1007/978-3-319-46478-7_31 10.1109/CVPR.2016.90 10.1109/CVPR.2014.220 10.1007/978-3-030-01231-1_17 10.1109/CVPR.2015.7298682 10.1109/CVPR42600.2020.00643 10.1007/978-3-030-01240-3_47 10.1109/WACV.2016.7477558 10.1109/CVPR.2015.7298594 10.1109/ICCV.2019.00659 10.1109/CVPR42600.2020.00525 10.1109/CVPR.2018.00552 10.1109/FG.2018.00020 10.1109/CVPR.2018.00745 10.1109/LSP.2018.2822810 10.1109/CVPR.2019.00482 10.1109/CVPR.2019.00516 10.1007/978-3-319-46487-9_6 10.1109/CVPR.2019.01108 10.5244/C.29.41 10.1109/CVPR.2018.00474 |
| ContentType | Journal Article |
| Copyright | 1225‐6463/$ © 2024 ETRI |
| Copyright_xml | – notice: 1225‐6463/$ © 2024 ETRI |
| DBID | AAYXX CITATION DOA ACYCR |
| DOI | 10.4218/etrij.2023-0167 |
| DatabaseName | CrossRef DOAJ Open Access Full Text Korean Citation Index |
| DatabaseTitle | CrossRef |
| DatabaseTitleList | CrossRef |
| Database_xml | – sequence: 1 dbid: DOA name: DOAJ Directory of Open Access Journals url: https://www.doaj.org/ sourceTypes: Open Website |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Engineering |
| EISSN | 2233-7326 |
| EndPage | 670 |
| ExternalDocumentID | oai_kci_go_kr_ARTI_10586349 oai_doaj_org_article_6e69ea2ca9ad4cda92cea694ec3274c8 10_4218_etrij_2023_0167 ETR212654 |
| Genre | article |
| GrantInformation_xml | – fundername: National Research Foundation of Korea funderid: Korea government(MSIT)/NRF‐2019R1G1A1003312; Korea government(MSIT)/NRF‐2021R1I1A3052815 |
| GroupedDBID | -~X .4S .DC 0R~ 29G 2WC 5GY 5VS 9ZL AAKPC AAMMB ACGFS ACXQS ACYCR ADBBV ADDVE ADMLS AEFGJ AENEX AGXDD AIDQK AIDYY ALMA_UNASSIGNED_HOLDINGS ARCSS AVUZU BCNDV DU5 E3Z EBS EDO EJD GROUPED_DOAJ IPNFZ ITG ITH JDI KQ8 KVFHK MK~ ML~ O9- OK1 OVT RIG RNS TR2 TUS WIN XSB AAYXX CITATION .UV 1OC P5Y |
| ID | FETCH-LOGICAL-c3844-a2036f541d90effbcd85ab2df79c6efe2bb62b8f570a53e5134a6e60e056192f3 |
| IEDL.DBID | DOA |
| ISSN | 1225-6463 |
| IngestDate | Fri Aug 16 03:20:37 EDT 2024 Wed Aug 27 01:06:38 EDT 2025 Tue Jul 01 02:03:21 EDT 2025 Sun Jul 06 04:45:32 EDT 2025 |
| IsDoiOpenAccess | true |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 4 |
| Language | English |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c3844-a2036f541d90effbcd85ab2df79c6efe2bb62b8f570a53e5134a6e60e056192f3 |
| Notes | https://doi.org/10.4218/etrij.2023-0167 |
| ORCID | 0000-0001-6530-5735 0000-0002-9193-8998 |
| OpenAccessLink | https://doaj.org/article/6e69ea2ca9ad4cda92cea694ec3274c8 |
| PageCount | 11 |
| ParticipantIDs | nrf_kci_oai_kci_go_kr_ARTI_10586349 doaj_primary_oai_doaj_org_article_6e69ea2ca9ad4cda92cea694ec3274c8 crossref_primary_10_4218_etrij_2023_0167 wiley_primary_10_4218_etrij_2023_0167_ETR212654 |
| ProviderPackageCode | CITATION AAYXX |
| PublicationCentury | 2000 |
| PublicationDate | August 2024 2024-08-00 2024-08-01 2024-08 |
| PublicationDateYYYYMMDD | 2024-08-01 |
| PublicationDate_xml | – month: 08 year: 2024 text: August 2024 |
| PublicationDecade | 2020 |
| PublicationTitle | ETRI journal |
| PublicationYear | 2024 |
| Publisher | Electronics and Telecommunications Research Institute (ETRI) 한국전자통신연구원 |
| Publisher_xml | – name: Electronics and Telecommunications Research Institute (ETRI) – name: 한국전자통신연구원 |
| References | 2021 2020 2019 2008 2018 2006 2017 2016 2005 2015 2014 2012; 13 2018; 25 e_1_2_9_31_1 e_1_2_9_11_1 e_1_2_9_10_1 e_1_2_9_13_1 e_1_2_9_32_1 e_1_2_9_12_1 e_1_2_9_15_1 e_1_2_9_14_1 e_1_2_9_17_1 e_1_2_9_16_1 e_1_2_9_19_1 e_1_2_9_18_1 e_1_2_9_20_1 e_1_2_9_22_1 e_1_2_9_21_1 e_1_2_9_24_1 e_1_2_9_23_1 e_1_2_9_8_1 e_1_2_9_7_1 e_1_2_9_6_1 e_1_2_9_5_1 e_1_2_9_4_1 e_1_2_9_3_1 e_1_2_9_2_1 Bergstra J. (e_1_2_9_30_1) 2012; 13 e_1_2_9_9_1 e_1_2_9_26_1 e_1_2_9_25_1 e_1_2_9_28_1 e_1_2_9_27_1 e_1_2_9_29_1 |
| References_xml | – start-page: 4690 year: 2019 end-page: 4699 – start-page: 770 year: 2016 end-page: 778 – start-page: 1735 year: 2006 end-page: 1742 – volume: 13 start-page: 281 issue: 2 year: 2012 end-page: 305 article-title: Random search for hyper‐parameter optimization publication-title: J. Mach. Learn. Res. – year: 2021 – start-page: 1701 year: 2014 end-page: 1708 – start-page: 765 year: 2018 end-page: 780 – start-page: 5022 year: 2019 end-page: 5030 – start-page: 539 year: 2005 end-page: 546 – start-page: 7132 year: 2018 end-page: 7141 – start-page: 6398 year: 2020 end-page: 6407 – year: 2014 – start-page: 1 year: 2016 end-page: 9 – start-page: 67 year: 2018 end-page: 74 – start-page: 815 year: 2015 end-page: 823 – start-page: 2821 year: 2017 end-page: 2829 – start-page: 1 year: 2015 end-page: 9 – year: 2008 – start-page: 87 year: 2016 end-page: 102 – start-page: 10823 year: 2019 end-page: 10832 – start-page: 499 year: 2016 end-page: 515 – start-page: 269 year: 2018 end-page: 285 – start-page: 6490 year: 2019 end-page: 6499 – start-page: 41.1 year: 2015 end-page: 41.12 – start-page: 212 year: 2017 end-page: 220 – start-page: 51 year: 2017 end-page: 59 – start-page: 5203 year: 2020 end-page: 5212 – start-page: 4510 year: 2018 end-page: 4520 – start-page: 5265 year: 2018 end-page: 5274 – start-page: 1857 year: 2016 end-page: 1865 – volume: 25 start-page: 926 issue: 7 year: 2018 end-page: 930 article-title: Additive margin softmax for face verification publication-title: IEEE Sig. Process. Lett. – ident: e_1_2_9_9_1 doi: 10.1109/CVPR.2005.202 – volume: 13 start-page: 281 issue: 2 year: 2012 ident: e_1_2_9_30_1 article-title: Random search for hyper‐parameter optimization publication-title: J. Mach. Learn. Res. – ident: e_1_2_9_10_1 doi: 10.1109/CVPR.2006.100 – ident: e_1_2_9_18_1 doi: 10.1109/CVPR.2017.713 – ident: e_1_2_9_22_1 doi: 10.1109/ICCV.2017.307 – ident: e_1_2_9_28_1 – ident: e_1_2_9_25_1 – ident: e_1_2_9_27_1 doi: 10.1109/CVPRW.2017.250 – ident: e_1_2_9_14_1 doi: 10.1007/978-3-319-46478-7_31 – ident: e_1_2_9_2_1 doi: 10.1109/CVPR.2016.90 – ident: e_1_2_9_13_1 doi: 10.1109/CVPR.2014.220 – ident: e_1_2_9_21_1 doi: 10.1007/978-3-030-01231-1_17 – ident: e_1_2_9_12_1 doi: 10.1109/CVPR.2015.7298682 – ident: e_1_2_9_29_1 doi: 10.1109/CVPR42600.2020.00643 – ident: e_1_2_9_6_1 doi: 10.1007/978-3-030-01240-3_47 – ident: e_1_2_9_4_1 – ident: e_1_2_9_26_1 doi: 10.1109/WACV.2016.7477558 – ident: e_1_2_9_5_1 doi: 10.1109/CVPR.2015.7298594 – ident: e_1_2_9_23_1 doi: 10.1109/ICCV.2019.00659 – ident: e_1_2_9_7_1 doi: 10.1109/CVPR42600.2020.00525 – ident: e_1_2_9_20_1 doi: 10.1109/CVPR.2018.00552 – ident: e_1_2_9_8_1 doi: 10.1109/FG.2018.00020 – ident: e_1_2_9_3_1 doi: 10.1109/CVPR.2018.00745 – ident: e_1_2_9_19_1 doi: 10.1109/LSP.2018.2822810 – ident: e_1_2_9_17_1 doi: 10.1109/CVPR.2019.00482 – ident: e_1_2_9_15_1 – ident: e_1_2_9_16_1 doi: 10.1109/CVPR.2019.00516 – ident: e_1_2_9_31_1 doi: 10.1007/978-3-319-46487-9_6 – ident: e_1_2_9_24_1 doi: 10.1109/CVPR.2019.01108 – ident: e_1_2_9_11_1 doi: 10.5244/C.29.41 – ident: e_1_2_9_32_1 doi: 10.1109/CVPR.2018.00474 |
| SSID | ssj0020458 |
| Score | 2.3565688 |
| Snippet | The performance of face recognition (FR) has reached a plateau for public benchmark datasets, such as labeled faces in the wild (LFW), celebrities in... The performance of face recognition (FR) has reached a plateau for publicbenchmark datasets, such as labeled faces in the wild (LFW), celebrities in... |
| SourceID | nrf doaj crossref wiley |
| SourceType | Open Website Index Database Publisher |
| StartPage | 660 |
| SubjectTerms | face recognition face verification 전자/정보통신공학 |
| Title | MixFace: Improving face verification with a focus on fine‐grained conditions |
| URI | https://onlinelibrary.wiley.com/doi/abs/10.4218%2Fetrij.2023-0167 https://doaj.org/article/6e69ea2ca9ad4cda92cea694ec3274c8 https://www.kci.go.kr/kciportal/ci/sereArticleSearch/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART003106932 |
| Volume | 46 |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| ispartofPNX | ETRI Journal, 2024, 46(4), , pp.660-670 |
| link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1LSwMxEA7iSQ_iE-uLgB68xO5mk3TjTcWioh6kBW8hT6mFVmoL_nxnsm3Rkxcvu2TZTcI3gZkvmf2GkDPlVVGnIjJeesUEh4tzUTKhKhdkp-j4XD_l6Vnd9cXDq3z9UeoLc8IaeeAGuLaKSkfLvdU2CB-s5j5apUX0FRAqn3_zLXSxIFNzqoXHf0i1YLUyBeM2oj4C_FkbC1W9X2DVcIYp-L_8UZbtBy8zmqTfwWr2Nt1NsjEPE-lVM70tshJH22T9h3jgDnl8Gnx1rY-XdLkxQBO0KSxOzP_JkFPcZ6WWprGffVJoJ-iAvWFdiBgocOHQpGztkn73tndzx-a1EZivaiGYxQPEJEUZdBFTcj7U0joeUkd7FVPkzinu6gRwW1lFWVbCApRFRMagear2yOpoPIr7hDoVIKbhtXIFqgWWjtepDNCx7YCr0qJFzhcImY9GAsMAdUAwTQbTIJgGwWyRa0Rw-RpqV-cHYFEzt6j5y6Itcgr4m6Ef5O_x_jY2w4mBCP8eRpa1qoRukXa2z19TMre9F3DPSoqD_5jcIVmDrkWTAnhEVqeTWTyGsGTqTvIK_AaoUd17 |
| linkProvider | Directory of Open Access Journals |
| linkToHtml | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LbxMxEB5BegAOiKealoclOHBZuuu1nXVvBTVKockBJajiYvkZhUoJWlKpx_4EfmN_CTPeELVcEJddeWXPWt94dh6eHQO8VV6VTSpjwSuvCsHx4lyUhVC1C3JQDnw-P2U8UaOZ-HQmz278C9PVh9gG3Egy8veaBJwC0iTlAtUScXHdLr6_p-O_C8qlvws7ZNvwHuwcfZ19m23dLtoKJLcLV26hcA5dgR8icvAXiVu6KZfwR42zbNNtwzVrnuEjeLgxGdlRx-PHcCcun8CDG4UEn8JkvLgcWh8P2TZIwBK2GS5UygXK8DOKuTLL0spf_GTYTkjg-urXnE6JiIGhZxy6BK5nMBseTz-Ois1JCYWvGyEKS9uJSYoq6DKm5HxopHU8pIH2KqbInVPcNQnBt7KOsqqFVVGVkfwHzVP9HHrL1TLuAnMqoIXDG-VKqh1YOd6kKiBhO0DFpUUf3v3ByPzoCmIYdCQITpPhNASnITj78IEw3HajStb5waqdm41gGJyGjpZ7q20QPljNfbRKi-hrdJh904c3yAFz7hd5PN3nK3PeGrT3T_DNslG10H04yBz615TM8fQLKmslxd5_j3gN90bT8ak5PZl83of72EN0mYAvoLduL-JLtE7W7tVm-f0GzHreLQ |
| linkToPdf | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1fTxQxEJ8gJEYejPgnHgg20QdfVna7bW_rGygXQLkYwxnjS9O_l5PkjixHwqMfwc_IJ2Gme1zEF8PLbrppZ5vfzOzMtLNTgLfKq7JJZSx45VUhOF6ci7IQqnZB9su-z-ennAzV4Ugc_5C32YT0L0xXH2K54Eaakb_XpODnIZGSC7RKxMR5O_n1nk7_LiiV_gGsUa08FPW1ve-jn6Nl1EU7gRR1oeAWCqfQ1fchIrv_kLhjmnIFfzQ40zbd9Vuz4Rk8gccLj5HtdSzegJU4fQrrf9URfAbDk8nVwPr4gS3XCFjCNkM5pVSgjD6jJVdmWZr5ywuG7YQErn__GdMhETEwDIxDl7_1HEaDg9OPh8XioITC140QhaXdxCRFFXQZU3I-NNI6HlJfexVT5M4p7pqE2FtZR1nVwqqoykjhg-apfgGr09k0vgTmVEAHhzfKlVQ6sHK8SVVAwraPdkuLHry7xcicd_UwDMYRBKfJcBqC0xCcPdgnDJfdqJB1fjBrx2ahFwanoaPl3mobhA9Wcx-t0iL6GuNl3_TgDXLAnPlJHk_38cyctQbd_SN8s2xULXQPdjOH_jclc3D6DW21kmLz3iNew8Ovnwbmy9Hw8xY8wg6iywN8Bavz9jJuo28ydzsL6bsBhhPdVg |
| 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=MixFace%3A+Improving+face+verification+with+a+focus+on+fine%E2%80%90grained+conditions&rft.jtitle=ETRI+journal&rft.au=Jung%2C+Junuk&rft.au=Son%2C+Sungbin&rft.au=Park%2C+Joochan&rft.au=Park%2C+Yongjun&rft.date=2024-08-01&rft.issn=1225-6463&rft.eissn=2233-7326&rft.volume=46&rft.issue=4&rft.spage=660&rft.epage=670&rft_id=info:doi/10.4218%2Fetrij.2023-0167&rft.externalDBID=10.4218%252Fetrij.2023-0167&rft.externalDocID=ETR212654 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1225-6463&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1225-6463&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1225-6463&client=summon |