LocMix: local saliency-based data augmentation for image classification

Data augmentation is a crucial strategy to tackle issues like inadequate model robustness and a significant generalization gap. It is proven to combat overfitting, elevate deep neural network performance, and enhance generalization, particularly when data are limited. In recent years, mixed sample d...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Signal, image and video processing image and video processing, 2024-03, Vol.18 (2), p.1383-1392
Hauptverfasser: Yan, Lingyu, Ye, Yu, Wang, Chunzhi, Sun, Yun
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 1392
container_issue 2
container_start_page 1383
container_title Signal, image and video processing
container_volume 18
creator Yan, Lingyu
Ye, Yu
Wang, Chunzhi
Sun, Yun
description Data augmentation is a crucial strategy to tackle issues like inadequate model robustness and a significant generalization gap. It is proven to combat overfitting, elevate deep neural network performance, and enhance generalization, particularly when data are limited. In recent years, mixed sample data augmentation (MSDA), including variants like Mixup and CutMix, has gained significant attention. However, these methods sometimes confound the network with misleading signals, limiting their effectiveness. In this context, we propose LocMix, an MSDA that aims to generate new training samples by prioritizing local saliency feature information and employing statistical data mixing. We achieve this by concealing salient regions with random masks and efficiently combining images through the optimization of local saliency information using transportation methods. Prioritizing the local features within an image allows LocMix to capture image details with greater accuracy and comprehensiveness, thereby enhancing the model’s capacity to understand the target image. We conduct extensive validation of this approach on various challenging datasets. When applied to the training of the PreAct-ResNet18 model, our method yields notable improvements in accuracy. Specifically, on the CIFAR-10 dataset, we observe an impressive 1.71% accuracy enhancement. Similarly, on CIFAR-100, Tiny-ImageNet, ImageNet, and SVHN, we attain substantial accuracy improvements of 80.12%, 64.60%, 77.62%, and 97.12%, corresponding to improvements of 4.88%, 8.75%, 1.93%, and 0.57%, respectively. These experimental results plainly illustrate the effectiveness of our proposed method.
doi_str_mv 10.1007/s11760-023-02852-0
format Article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_journals_2928896393</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2928896393</sourcerecordid><originalsourceid>FETCH-LOGICAL-c319t-985d43c65506b0591464d56a4a9e3e0336fc5483172895df26275a8e6f93ae393</originalsourceid><addsrcrecordid>eNp9kE1LAzEQhoMoWGr_gKeA52g-NtnEmxRthYoXPYdpNilbtpuabMH-e2NX9ObAMAPzvjPDg9A1o7eM0vouM1YrSigXJbXkhJ6hCdNKEFYzdv7bU3GJZjlvaQnBa630BC1W0b20n_e4iw46nKFrfe-OZA3ZN7iBATAcNjvfDzC0scchJtzuYOOx6yDnNrTuNLhCFwG67Gc_dYrenx7f5kuyel08zx9WxAlmBmK0bCrhlJRUrak0rFJVIxVUYLzwVAgVnKy0YDXXRjaBK15L0F4FI8ALI6boZty7T_Hj4PNgt_GQ-nLScsO1NqqIioqPKpdizskHu0_l63S0jNpvZnZkZgsze2JmaTGJ0ZSLuN_49Lf6H9cXVXRs_A</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2928896393</pqid></control><display><type>article</type><title>LocMix: local saliency-based data augmentation for image classification</title><source>SpringerLink Journals - AutoHoldings</source><creator>Yan, Lingyu ; Ye, Yu ; Wang, Chunzhi ; Sun, Yun</creator><creatorcontrib>Yan, Lingyu ; Ye, Yu ; Wang, Chunzhi ; Sun, Yun</creatorcontrib><description>Data augmentation is a crucial strategy to tackle issues like inadequate model robustness and a significant generalization gap. It is proven to combat overfitting, elevate deep neural network performance, and enhance generalization, particularly when data are limited. In recent years, mixed sample data augmentation (MSDA), including variants like Mixup and CutMix, has gained significant attention. However, these methods sometimes confound the network with misleading signals, limiting their effectiveness. In this context, we propose LocMix, an MSDA that aims to generate new training samples by prioritizing local saliency feature information and employing statistical data mixing. We achieve this by concealing salient regions with random masks and efficiently combining images through the optimization of local saliency information using transportation methods. Prioritizing the local features within an image allows LocMix to capture image details with greater accuracy and comprehensiveness, thereby enhancing the model’s capacity to understand the target image. We conduct extensive validation of this approach on various challenging datasets. When applied to the training of the PreAct-ResNet18 model, our method yields notable improvements in accuracy. Specifically, on the CIFAR-10 dataset, we observe an impressive 1.71% accuracy enhancement. Similarly, on CIFAR-100, Tiny-ImageNet, ImageNet, and SVHN, we attain substantial accuracy improvements of 80.12%, 64.60%, 77.62%, and 97.12%, corresponding to improvements of 4.88%, 8.75%, 1.93%, and 0.57%, respectively. These experimental results plainly illustrate the effectiveness of our proposed method.</description><identifier>ISSN: 1863-1703</identifier><identifier>EISSN: 1863-1711</identifier><identifier>DOI: 10.1007/s11760-023-02852-0</identifier><language>eng</language><publisher>London: Springer London</publisher><subject>Accuracy ; Artificial neural networks ; Computer Imaging ; Computer Science ; Data augmentation ; Datasets ; Effectiveness ; Image classification ; Image enhancement ; Image Processing and Computer Vision ; Multimedia Information Systems ; Original Paper ; Pattern Recognition and Graphics ; Salience ; Signal,Image and Speech Processing ; Vision</subject><ispartof>Signal, image and video processing, 2024-03, Vol.18 (2), p.1383-1392</ispartof><rights>The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2023. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c319t-985d43c65506b0591464d56a4a9e3e0336fc5483172895df26275a8e6f93ae393</citedby><cites>FETCH-LOGICAL-c319t-985d43c65506b0591464d56a4a9e3e0336fc5483172895df26275a8e6f93ae393</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://link.springer.com/content/pdf/10.1007/s11760-023-02852-0$$EPDF$$P50$$Gspringer$$H</linktopdf><linktohtml>$$Uhttps://link.springer.com/10.1007/s11760-023-02852-0$$EHTML$$P50$$Gspringer$$H</linktohtml><link.rule.ids>314,780,784,27923,27924,41487,42556,51318</link.rule.ids></links><search><creatorcontrib>Yan, Lingyu</creatorcontrib><creatorcontrib>Ye, Yu</creatorcontrib><creatorcontrib>Wang, Chunzhi</creatorcontrib><creatorcontrib>Sun, Yun</creatorcontrib><title>LocMix: local saliency-based data augmentation for image classification</title><title>Signal, image and video processing</title><addtitle>SIViP</addtitle><description>Data augmentation is a crucial strategy to tackle issues like inadequate model robustness and a significant generalization gap. It is proven to combat overfitting, elevate deep neural network performance, and enhance generalization, particularly when data are limited. In recent years, mixed sample data augmentation (MSDA), including variants like Mixup and CutMix, has gained significant attention. However, these methods sometimes confound the network with misleading signals, limiting their effectiveness. In this context, we propose LocMix, an MSDA that aims to generate new training samples by prioritizing local saliency feature information and employing statistical data mixing. We achieve this by concealing salient regions with random masks and efficiently combining images through the optimization of local saliency information using transportation methods. Prioritizing the local features within an image allows LocMix to capture image details with greater accuracy and comprehensiveness, thereby enhancing the model’s capacity to understand the target image. We conduct extensive validation of this approach on various challenging datasets. When applied to the training of the PreAct-ResNet18 model, our method yields notable improvements in accuracy. Specifically, on the CIFAR-10 dataset, we observe an impressive 1.71% accuracy enhancement. Similarly, on CIFAR-100, Tiny-ImageNet, ImageNet, and SVHN, we attain substantial accuracy improvements of 80.12%, 64.60%, 77.62%, and 97.12%, corresponding to improvements of 4.88%, 8.75%, 1.93%, and 0.57%, respectively. These experimental results plainly illustrate the effectiveness of our proposed method.</description><subject>Accuracy</subject><subject>Artificial neural networks</subject><subject>Computer Imaging</subject><subject>Computer Science</subject><subject>Data augmentation</subject><subject>Datasets</subject><subject>Effectiveness</subject><subject>Image classification</subject><subject>Image enhancement</subject><subject>Image Processing and Computer Vision</subject><subject>Multimedia Information Systems</subject><subject>Original Paper</subject><subject>Pattern Recognition and Graphics</subject><subject>Salience</subject><subject>Signal,Image and Speech Processing</subject><subject>Vision</subject><issn>1863-1703</issn><issn>1863-1711</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNp9kE1LAzEQhoMoWGr_gKeA52g-NtnEmxRthYoXPYdpNilbtpuabMH-e2NX9ObAMAPzvjPDg9A1o7eM0vouM1YrSigXJbXkhJ6hCdNKEFYzdv7bU3GJZjlvaQnBa630BC1W0b20n_e4iw46nKFrfe-OZA3ZN7iBATAcNjvfDzC0scchJtzuYOOx6yDnNrTuNLhCFwG67Gc_dYrenx7f5kuyel08zx9WxAlmBmK0bCrhlJRUrak0rFJVIxVUYLzwVAgVnKy0YDXXRjaBK15L0F4FI8ALI6boZty7T_Hj4PNgt_GQ-nLScsO1NqqIioqPKpdizskHu0_l63S0jNpvZnZkZgsze2JmaTGJ0ZSLuN_49Lf6H9cXVXRs_A</recordid><startdate>20240301</startdate><enddate>20240301</enddate><creator>Yan, Lingyu</creator><creator>Ye, Yu</creator><creator>Wang, Chunzhi</creator><creator>Sun, Yun</creator><general>Springer London</general><general>Springer Nature B.V</general><scope>AAYXX</scope><scope>CITATION</scope></search><sort><creationdate>20240301</creationdate><title>LocMix: local saliency-based data augmentation for image classification</title><author>Yan, Lingyu ; Ye, Yu ; Wang, Chunzhi ; Sun, Yun</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c319t-985d43c65506b0591464d56a4a9e3e0336fc5483172895df26275a8e6f93ae393</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Accuracy</topic><topic>Artificial neural networks</topic><topic>Computer Imaging</topic><topic>Computer Science</topic><topic>Data augmentation</topic><topic>Datasets</topic><topic>Effectiveness</topic><topic>Image classification</topic><topic>Image enhancement</topic><topic>Image Processing and Computer Vision</topic><topic>Multimedia Information Systems</topic><topic>Original Paper</topic><topic>Pattern Recognition and Graphics</topic><topic>Salience</topic><topic>Signal,Image and Speech Processing</topic><topic>Vision</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Yan, Lingyu</creatorcontrib><creatorcontrib>Ye, Yu</creatorcontrib><creatorcontrib>Wang, Chunzhi</creatorcontrib><creatorcontrib>Sun, Yun</creatorcontrib><collection>CrossRef</collection><jtitle>Signal, image and video processing</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Yan, Lingyu</au><au>Ye, Yu</au><au>Wang, Chunzhi</au><au>Sun, Yun</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>LocMix: local saliency-based data augmentation for image classification</atitle><jtitle>Signal, image and video processing</jtitle><stitle>SIViP</stitle><date>2024-03-01</date><risdate>2024</risdate><volume>18</volume><issue>2</issue><spage>1383</spage><epage>1392</epage><pages>1383-1392</pages><issn>1863-1703</issn><eissn>1863-1711</eissn><abstract>Data augmentation is a crucial strategy to tackle issues like inadequate model robustness and a significant generalization gap. It is proven to combat overfitting, elevate deep neural network performance, and enhance generalization, particularly when data are limited. In recent years, mixed sample data augmentation (MSDA), including variants like Mixup and CutMix, has gained significant attention. However, these methods sometimes confound the network with misleading signals, limiting their effectiveness. In this context, we propose LocMix, an MSDA that aims to generate new training samples by prioritizing local saliency feature information and employing statistical data mixing. We achieve this by concealing salient regions with random masks and efficiently combining images through the optimization of local saliency information using transportation methods. Prioritizing the local features within an image allows LocMix to capture image details with greater accuracy and comprehensiveness, thereby enhancing the model’s capacity to understand the target image. We conduct extensive validation of this approach on various challenging datasets. When applied to the training of the PreAct-ResNet18 model, our method yields notable improvements in accuracy. Specifically, on the CIFAR-10 dataset, we observe an impressive 1.71% accuracy enhancement. Similarly, on CIFAR-100, Tiny-ImageNet, ImageNet, and SVHN, we attain substantial accuracy improvements of 80.12%, 64.60%, 77.62%, and 97.12%, corresponding to improvements of 4.88%, 8.75%, 1.93%, and 0.57%, respectively. These experimental results plainly illustrate the effectiveness of our proposed method.</abstract><cop>London</cop><pub>Springer London</pub><doi>10.1007/s11760-023-02852-0</doi><tpages>10</tpages></addata></record>
fulltext fulltext
identifier ISSN: 1863-1703
ispartof Signal, image and video processing, 2024-03, Vol.18 (2), p.1383-1392
issn 1863-1703
1863-1711
language eng
recordid cdi_proquest_journals_2928896393
source SpringerLink Journals - AutoHoldings
subjects Accuracy
Artificial neural networks
Computer Imaging
Computer Science
Data augmentation
Datasets
Effectiveness
Image classification
Image enhancement
Image Processing and Computer Vision
Multimedia Information Systems
Original Paper
Pattern Recognition and Graphics
Salience
Signal,Image and Speech Processing
Vision
title LocMix: local saliency-based data augmentation for image classification
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-12T00%3A02%3A07IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=LocMix:%20local%20saliency-based%20data%20augmentation%20for%20image%20classification&rft.jtitle=Signal,%20image%20and%20video%20processing&rft.au=Yan,%20Lingyu&rft.date=2024-03-01&rft.volume=18&rft.issue=2&rft.spage=1383&rft.epage=1392&rft.pages=1383-1392&rft.issn=1863-1703&rft.eissn=1863-1711&rft_id=info:doi/10.1007/s11760-023-02852-0&rft_dat=%3Cproquest_cross%3E2928896393%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2928896393&rft_id=info:pmid/&rfr_iscdi=true