Loading…
Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation
Although Convolutional Neural Networks (CNNs) have high accuracy in image recognition, they are vulnerable to adversarial examples and out-of-distribution data, and the difference from human recognition has been pointed out. In order to improve the robustness against out-of-distribution data, we pre...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Conference Proceeding |
| Language: | English |
| Subjects: | |
| Online Access: | Request full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| cited_by | |
|---|---|
| cites | |
| container_end_page | 3120 |
| container_issue | |
| container_start_page | 3116 |
| container_title | |
| container_volume | |
| creator | Mukai, Koki Kumano, Soichiro Yamasaki, Toshihiko |
| description | Although Convolutional Neural Networks (CNNs) have high accuracy in image recognition, they are vulnerable to adversarial examples and out-of-distribution data, and the difference from human recognition has been pointed out. In order to improve the robustness against out-of-distribution data, we present a frequency-based data augmentation technique that replaces the frequency components with other images of the same class. When the training data are CIFAR10 and the out-of-distribution data are SVHN, the Area Under Receiver Operating Characteristic (AUROC) curve of the model trained with the proposed method increases from 89.22% to 98.15%, and further increased to 98.59% when combined with another data augmentation method. Furthermore, we experimentally demonstrate that the robust model for out-of-distribution data uses a lot of high-frequency components of the image. |
| doi_str_mv | 10.1109/ICIP46576.2022.9897504 |
| format | conference_proceeding |
| fullrecord | <record><control><sourceid>ieee_CHZPO</sourceid><recordid>TN_cdi_ieee_primary_9897504</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>9897504</ieee_id><sourcerecordid>9897504</sourcerecordid><originalsourceid>FETCH-LOGICAL-i203t-c44bf798ffc8378b6b8adb6386e384508db123cf564229a3354efbc92a2c88e73</originalsourceid><addsrcrecordid>eNotj91KwzAARqMguE2fQJC8QGr-fy5n57QwcIhejyRNRsS2s0mFvb0Td_Vx4HDgA-Ce4IoQbB6autlyKZSsKKa0MtoogfkFmBMpBTeSYnUJZpRpgvSJr8E850-MKSaMzMC26Q7j8JP6PXwb3JRLH3KGZYDDVNAQ0SrlMiY3lTT0cGWLhe4I12P4nkLvj-jR5tDC5bTvQl_sn3QDrqL9yuH2vAvwsX56r1_Q5vW5qZcblChmBXnOXVRGx-g1U9pJp23rJNMyMM0F1q0jlPkoJKfUWMYED9F5Qy31WgfFFuDuv5tCCLvDmDo7Hnfn9-wXRURQTQ</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation</title><source>IEEE Xplore All Conference Series</source><creator>Mukai, Koki ; Kumano, Soichiro ; Yamasaki, Toshihiko</creator><creatorcontrib>Mukai, Koki ; Kumano, Soichiro ; Yamasaki, Toshihiko</creatorcontrib><description>Although Convolutional Neural Networks (CNNs) have high accuracy in image recognition, they are vulnerable to adversarial examples and out-of-distribution data, and the difference from human recognition has been pointed out. In order to improve the robustness against out-of-distribution data, we present a frequency-based data augmentation technique that replaces the frequency components with other images of the same class. When the training data are CIFAR10 and the out-of-distribution data are SVHN, the Area Under Receiver Operating Characteristic (AUROC) curve of the model trained with the proposed method increases from 89.22% to 98.15%, and further increased to 98.59% when combined with another data augmentation method. Furthermore, we experimentally demonstrate that the robust model for out-of-distribution data uses a lot of high-frequency components of the image.</description><identifier>EISSN: 2381-8549</identifier><identifier>EISBN: 1665496207</identifier><identifier>EISBN: 9781665496209</identifier><identifier>DOI: 10.1109/ICIP46576.2022.9897504</identifier><language>eng</language><publisher>IEEE</publisher><subject>Convolutional neural networks ; data augmentation ; Data models ; frequency ; Image recognition ; neural network ; out-of-distribution ; Receivers ; Robustness ; Training data</subject><ispartof>2022 IEEE International Conference on Image Processing (ICIP), 2022, p.3116-3120</ispartof><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/9897504$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>309,310,780,784,789,790,23930,23931,25140,27925,54555,54932</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/9897504$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Mukai, Koki</creatorcontrib><creatorcontrib>Kumano, Soichiro</creatorcontrib><creatorcontrib>Yamasaki, Toshihiko</creatorcontrib><title>Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation</title><title>2022 IEEE International Conference on Image Processing (ICIP)</title><addtitle>ICIP</addtitle><description>Although Convolutional Neural Networks (CNNs) have high accuracy in image recognition, they are vulnerable to adversarial examples and out-of-distribution data, and the difference from human recognition has been pointed out. In order to improve the robustness against out-of-distribution data, we present a frequency-based data augmentation technique that replaces the frequency components with other images of the same class. When the training data are CIFAR10 and the out-of-distribution data are SVHN, the Area Under Receiver Operating Characteristic (AUROC) curve of the model trained with the proposed method increases from 89.22% to 98.15%, and further increased to 98.59% when combined with another data augmentation method. Furthermore, we experimentally demonstrate that the robust model for out-of-distribution data uses a lot of high-frequency components of the image.</description><subject>Convolutional neural networks</subject><subject>data augmentation</subject><subject>Data models</subject><subject>frequency</subject><subject>Image recognition</subject><subject>neural network</subject><subject>out-of-distribution</subject><subject>Receivers</subject><subject>Robustness</subject><subject>Training data</subject><issn>2381-8549</issn><isbn>1665496207</isbn><isbn>9781665496209</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2022</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNotj91KwzAARqMguE2fQJC8QGr-fy5n57QwcIhejyRNRsS2s0mFvb0Td_Vx4HDgA-Ce4IoQbB6autlyKZSsKKa0MtoogfkFmBMpBTeSYnUJZpRpgvSJr8E850-MKSaMzMC26Q7j8JP6PXwb3JRLH3KGZYDDVNAQ0SrlMiY3lTT0cGWLhe4I12P4nkLvj-jR5tDC5bTvQl_sn3QDrqL9yuH2vAvwsX56r1_Q5vW5qZcblChmBXnOXVRGx-g1U9pJp23rJNMyMM0F1q0jlPkoJKfUWMYED9F5Qy31WgfFFuDuv5tCCLvDmDo7Hnfn9-wXRURQTQ</recordid><startdate>20221016</startdate><enddate>20221016</enddate><creator>Mukai, Koki</creator><creator>Kumano, Soichiro</creator><creator>Yamasaki, Toshihiko</creator><general>IEEE</general><scope>6IE</scope><scope>6IH</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIO</scope></search><sort><creationdate>20221016</creationdate><title>Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation</title><author>Mukai, Koki ; Kumano, Soichiro ; Yamasaki, Toshihiko</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-i203t-c44bf798ffc8378b6b8adb6386e384508db123cf564229a3354efbc92a2c88e73</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Convolutional neural networks</topic><topic>data augmentation</topic><topic>Data models</topic><topic>frequency</topic><topic>Image recognition</topic><topic>neural network</topic><topic>out-of-distribution</topic><topic>Receivers</topic><topic>Robustness</topic><topic>Training data</topic><toplevel>online_resources</toplevel><creatorcontrib>Mukai, Koki</creatorcontrib><creatorcontrib>Kumano, Soichiro</creatorcontrib><creatorcontrib>Yamasaki, Toshihiko</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan (POP) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE/IET Electronic Library (IEL)</collection><collection>IEEE Proceedings Order Plans (POP) 1998-present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Mukai, Koki</au><au>Kumano, Soichiro</au><au>Yamasaki, Toshihiko</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation</atitle><btitle>2022 IEEE International Conference on Image Processing (ICIP)</btitle><stitle>ICIP</stitle><date>2022-10-16</date><risdate>2022</risdate><spage>3116</spage><epage>3120</epage><pages>3116-3120</pages><eissn>2381-8549</eissn><eisbn>1665496207</eisbn><eisbn>9781665496209</eisbn><abstract>Although Convolutional Neural Networks (CNNs) have high accuracy in image recognition, they are vulnerable to adversarial examples and out-of-distribution data, and the difference from human recognition has been pointed out. In order to improve the robustness against out-of-distribution data, we present a frequency-based data augmentation technique that replaces the frequency components with other images of the same class. When the training data are CIFAR10 and the out-of-distribution data are SVHN, the Area Under Receiver Operating Characteristic (AUROC) curve of the model trained with the proposed method increases from 89.22% to 98.15%, and further increased to 98.59% when combined with another data augmentation method. Furthermore, we experimentally demonstrate that the robust model for out-of-distribution data uses a lot of high-frequency components of the image.</abstract><pub>IEEE</pub><doi>10.1109/ICIP46576.2022.9897504</doi><tpages>5</tpages></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | EISSN: 2381-8549 |
| ispartof | 2022 IEEE International Conference on Image Processing (ICIP), 2022, p.3116-3120 |
| issn | 2381-8549 |
| language | eng |
| recordid | cdi_ieee_primary_9897504 |
| source | IEEE Xplore All Conference Series |
| subjects | Convolutional neural networks data augmentation Data models frequency Image recognition neural network out-of-distribution Receivers Robustness Training data |
| title | Improving Robustness to out-of-Distribution Data by Frequency-Based Augmentation |
| url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-08T00%3A30%3A35IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_CHZPO&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=Improving%20Robustness%20to%20out-of-Distribution%20Data%20by%20Frequency-Based%20Augmentation&rft.btitle=2022%20IEEE%20International%20Conference%20on%20Image%20Processing%20(ICIP)&rft.au=Mukai,%20Koki&rft.date=2022-10-16&rft.spage=3116&rft.epage=3120&rft.pages=3116-3120&rft.eissn=2381-8549&rft_id=info:doi/10.1109/ICIP46576.2022.9897504&rft.eisbn=1665496207&rft.eisbn_list=9781665496209&rft_dat=%3Cieee_CHZPO%3E9897504%3C/ieee_CHZPO%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-i203t-c44bf798ffc8378b6b8adb6386e384508db123cf564229a3354efbc92a2c88e73%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=9897504&rfr_iscdi=true |