{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T20:50:15Z","timestamp":1792788615962,"version":"4.5.0"},"reference-count":51,"publisher":"Public Library of Science (PLoS)","issue":"9","license":[{"start":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T00:00:00Z","timestamp":1790553600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Research Platform of Anhui Vocational and Technical University","award":["No. 2024xjpt04"],"award-info":[{"award-number":["No. 2024xjpt04"]}],"id":[{"id":"https:\/\/ror.org\/029razw56","id-type":"ROR","asserted-by":"crossref"}]},{"name":"Key Natural Science Research Project of Anhui Provincial Universities","award":["No. 2022AH052062"],"award-info":[{"award-number":["No. 2022AH052062"]}]}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"<jats:p>Few-shot industrial surface defect detection remains challenging because annotated defect images are scarce, defect scales vary considerably, and textured backgrounds can cause false responses. This study proposes a Prototype-Aligned Multi-Scale Attention Network (PA-MSAN) that integrates multi-receptive-field feature extraction, spatial-channel attention, and region of interest (RoI)-level prototype matching within a two-stage detector. Experiments are conducted on NEU-DET under 1-shot, 3-shot, and 5-shot episode settings, with 5-shot cross-domain evaluation on GC10-DET using target-domain support prototypes. Multi-scale feature-pyramid fusion represents local textures and broader defect morphology, while attention weighting enhances defect-sensitive regions and channels before region proposal network (RPN) proposal generation and RoI Align. Few-shot classification is performed by cosine similarity between RoI features and class prototypes, and training jointly optimizes detection classification, bounding-box regression, region proposal, and Prototype Alignment Loss (PAL). Under the 5-shot NEU-DET setting, PA-MSAN achieves 92.2\u2009\u00b1\u20090.4% mean average precision (mAP)@0.5, 88.7\u2009\u00b1\u20090.4% small-object average precision (APsmall), 0.89\u2009\u00b1\u20090.01 average intersection over union (IoU), and a 4.1\u2009\u00b1\u20090.2% background false response rate, with 23.4 ms detection time and 18.7\u2009\u00d7\u200910\u2076 parameters. Compared with Faster R-CNN, PA-MSAN improves mAP@0.5 by 2.9 percentage points and APsmall by 3.9 percentage points while reducing the background false response rate by 1.7 percentage points. Under the 5-shot target-domain support setting, PA-MSAN achieves 86.4\u2009\u00b1\u20090.4% mAP@0.5 on GC10-DET, 5.8 percentage points (6.3%) below the corresponding NEU-DET result. The results support region-level prototype alignment for limited-label metallic surface inspection under the evaluated datasets and two-stage configuration.<\/jats:p>","DOI":"10.1371\/journal.pone.0359400","type":"journal-article","created":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T17:29:02Z","timestamp":1790616542000},"page":"e0359400","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["A prototype-aligned multi-scale attention network for few-shot industrial surface defect detection"],"prefix":"10.1371","volume":"21","author":[{"given":"Rubing","family":"Huang","sequence":"first","affiliation":[{"name":"Anhui Vocational and Technical College","place":["Hefei, Anhui, China"],"department":["School of Computer and Information Technology"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-2019-7349","authenticated-orcid":true,"given":"Lianyi","family":"Zhao","sequence":"additional","affiliation":[{"name":"Anhui Water Conservancy and Hydropower Vocational and Technical College","place":["Hefei, Anhui, China"],"department":["School of Electrical and Mechanical Engineering"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,9,28]]},"reference":[{"issue":"4","key":"pone.0359400.ref001","doi-asserted-by":"crossref","first-page":"6463","DOI":"10.3233\/JIFS-223041","article-title":"Research on surface defects detection method and system in manufacturing processes based on the fusion of multi-scale features and semantic segmentation for intelligent manufacturing","volume":"44","author":"Y Gong","year":"2023","journal-title":"IFS"},{"key":"pone.0359400.ref002","doi-asserted-by":"crossref","first-page":"103901","DOI":"10.1016\/j.compind.2023.103901","article-title":"Few-shot semantic segmentation for industrial defect recognition","volume":"148","author":"X Shi","year":"2023","journal-title":"Comput Ind"},{"key":"pone.0359400.ref003","doi-asserted-by":"crossref","first-page":"111265","DOI":"10.1016\/j.knosys.2023.111265","article-title":"Few-shot defect recognition for the multi-domain industry via attention embedding and fine-grained feature enhancement","volume":"284","author":"Y Su","year":"2024","journal-title":"Knowledge-Based Systems"},{"issue":"10","key":"pone.0359400.ref004","doi-asserted-by":"crossref","first-page":"1727","DOI":"10.2355\/isijinternational.ISIJINT-2023-118","article-title":"Few-shot steel plate surface defect detection with multi-relation aggregation and adaptive support learning","volume":"63","author":"Y Deng","year":"2023","journal-title":"ISIJ Int"},{"issue":"7","key":"pone.0359400.ref005","doi-asserted-by":"crossref","first-page":"1662","DOI":"10.3390\/electronics12071662","article-title":"AMFF-YOLOX: Towards an attention mechanism and multiple feature fusion based on YOLOX for industrial defect detection","volume":"12","author":"Y Chen","year":"2023","journal-title":"Electronics"},{"key":"pone.0359400.ref006","doi-asserted-by":"crossref","first-page":"82","DOI":"10.1016\/j.patrec.2023.06.010","article-title":"Multi-scale attention and dilation network for small defect detection","volume":"172","author":"X Xiang","year":"2023","journal-title":"Pattern Recognition Letters"},{"key":"pone.0359400.ref007","doi-asserted-by":"crossref","first-page":"106368","DOI":"10.1016\/j.engappai.2023.106368","article-title":"Fine coordinate attention for surface defect detection","volume":"123","author":"M Xiao","year":"2023","journal-title":"Eng Appl Artif Intell"},{"key":"pone.0359400.ref008","doi-asserted-by":"crossref","first-page":"107134","DOI":"10.1016\/j.engappai.2023.107134","article-title":"ICA-Net: Industrial defect detection network based on convolutional attention guidance and aggregation of multiscale features","volume":"126","author":"S Zhao","year":"2023","journal-title":"Eng Appl Artif Intell"},{"issue":"15","key":"pone.0359400.ref009","doi-asserted-by":"crossref","first-page":"6982","DOI":"10.3390\/s23156982","article-title":"Defect detection in steel using a hybrid attention network","volume":"23","author":"M Zhou","year":"2023","journal-title":"Sensors (Basel)"},{"issue":"8","key":"pone.0359400.ref010","doi-asserted-by":"crossref","first-page":"1856","DOI":"10.3390\/electronics12081856","article-title":"Surface defect detection of hot rolled steel based on attention mechanism and dilated convolution for industrial robots","volume":"12","author":"Y Yu","year":"2023","journal-title":"Electronics"},{"key":"pone.0359400.ref011","doi-asserted-by":"crossref","first-page":"106369","DOI":"10.1016\/j.engappai.2023.106369","article-title":"Enriched multi-scale cascade pyramid features and guided context attention network for industrial surface defect detection","volume":"123","author":"L Shao","year":"2023","journal-title":"Eng Appl Artif Intell"},{"issue":"1","key":"pone.0359400.ref012","doi-asserted-by":"crossref","first-page":"88","DOI":"10.1049\/ipr2.12935","article-title":"FSSDD: Few\u2010shot steel defect detection based on multi\u2010scale semantic enhancement representation and mask category information mapping","volume":"18","author":"Z Liu","year":"2023","journal-title":"IET Image Processing"},{"key":"pone.0359400.ref013","doi-asserted-by":"crossref","first-page":"37813","DOI":"10.1109\/ACCESS.2026.3671458","article-title":"A few-shot surface defect detection method for industrial vision via bi-directional feature fusion and multi-stage refinement","volume":"14","author":"Y Zhong","year":"2026","journal-title":"IEEE Access"},{"key":"pone.0359400.ref014","doi-asserted-by":"crossref","first-page":"122365","DOI":"10.1016\/j.eswa.2023.122365","article-title":"A hierarchical attention detector for bearing surface defect detection","volume":"239","author":"J Ma","year":"2024","journal-title":"Expert Systems with Applications"},{"issue":"9","key":"pone.0359400.ref015","doi-asserted-by":"crossref","first-page":"9713","DOI":"10.1109\/TII.2023.3234030","article-title":"Attention-guided multitask learning for surface defect identification","volume":"19","author":"V Sampath","year":"2023","journal-title":"IEEE Trans Ind Inf"},{"key":"pone.0359400.ref016","doi-asserted-by":"crossref","first-page":"452","DOI":"10.1007\/s00530-025-02030-x","article-title":"Attention-guided few-shot learning for metal surface defect classification","volume":"31","author":"X Liu","year":"2025","journal-title":"Multimed Syst"},{"key":"pone.0359400.ref017","doi-asserted-by":"crossref","first-page":"109170","DOI":"10.1016\/j.patcog.2022.109170","article-title":"SAPENet: Self-attention based prototype enhancement network for few-shot learning","volume":"135","author":"X Huang","year":"2023","journal-title":"Pattern Recognition"},{"issue":"4","key":"pone.0359400.ref018","doi-asserted-by":"crossref","first-page":"2924","DOI":"10.1109\/TPAMI.2025.3529038","article-title":"Few-shot class-incremental learning for classification and object detection: A survey","volume":"47","author":"J Zhang","year":"2025","journal-title":"IEEE Trans Pattern Anal Mach Intell"},{"key":"pone.0359400.ref019","first-page":"9919","article-title":"Frustratingly simple few-shot object detection","volume":"119","author":"X Wang","year":"2020","journal-title":"Proceedings of Machine Learning Research"},{"key":"pone.0359400.ref020","doi-asserted-by":"crossref","unstructured":"Kang B, Liu Z, Wang X, Yu F, Feng J, Darrell T. Few-Shot Object Detection via Feature Reweighting. In: 2019 IEEE\/CVF International Conference on Computer Vision (ICCV), 2019. 8419\u201328. https:\/\/doi.org\/10.1109\/iccv.2019.00851","DOI":"10.1109\/ICCV.2019.00851"},{"key":"pone.0359400.ref021","doi-asserted-by":"crossref","unstructured":"Yan X, Chen Z, Xu A, Wang X, Liang X, Lin L. Meta R-CNN: Towards General Solver for Instance-Level Low-Shot Learning. In: 2019 IEEE\/CVF International Conference on Computer Vision (ICCV), 2019. 9576\u201385. https:\/\/doi.org\/10.1109\/iccv.2019.00967","DOI":"10.1109\/ICCV.2019.00967"},{"key":"pone.0359400.ref022","doi-asserted-by":"crossref","first-page":"858","DOI":"10.1016\/j.apsusc.2013.09.002","article-title":"A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects","volume":"285","author":"K Song","year":"2013","journal-title":"Appl Surf Sci"},{"issue":"4","key":"pone.0359400.ref023","doi-asserted-by":"crossref","first-page":"1493","DOI":"10.1109\/TIM.2019.2915404","article-title":"An end-to-end steel surface defect detection approach via fusing multiple hierarchical features","volume":"69","author":"Y He","year":"2020","journal-title":"IEEE Trans Instrum Meas"},{"issue":"1","key":"pone.0359400.ref024","doi-asserted-by":"crossref","first-page":"22136","DOI":"10.1038\/s41598-023-49359-9","article-title":"Surface defect detection of industrial components based on vision","volume":"13","author":"Z Chen","year":"2023","journal-title":"Sci Rep"},{"key":"pone.0359400.ref025","doi-asserted-by":"crossref","first-page":"102205","DOI":"10.1016\/j.aei.2023.102205","article-title":"ALADA: A lite automatic data augmentation framework for industrial defect detection","volume":"58","author":"Y Wang","year":"2023","journal-title":"Adv Eng Inform"},{"key":"pone.0359400.ref026","doi-asserted-by":"crossref","first-page":"43370","DOI":"10.1109\/ACCESS.2023.3271748","article-title":"Deep learning for automatic vision-based recognition of industrial surface defects: A survey","volume":"11","author":"M Prunella","year":"2023","journal-title":"IEEE Access"},{"issue":"6","key":"pone.0359400.ref027","doi-asserted-by":"crossref","first-page":"1562","DOI":"10.3390\/s20061562","article-title":"Deep metallic surface defect detection: the new benchmark and detection network","volume":"20","author":"X Lv","year":"2020","journal-title":"Sensors (Basel)"},{"issue":"3","key":"pone.0359400.ref028","doi-asserted-by":"crossref","first-page":"1141","DOI":"10.1007\/s10845-023-02097-1","article-title":"Hierarchical multi-scale network for cross-scale visual defect detection","volume":"35","author":"R Tang","year":"2024","journal-title":"J Intell Manuf"},{"issue":"2","key":"pone.0359400.ref029","doi-asserted-by":"crossref","first-page":"1054","DOI":"10.3390\/app13021054","article-title":"Surface defect detection of steel strip with double pyramid network","volume":"13","author":"X Zhou","year":"2023","journal-title":"Applied Sciences"},{"issue":"9","key":"pone.0359400.ref030","doi-asserted-by":"crossref","first-page":"4508","DOI":"10.3390\/s23094508","article-title":"HA-FPN: Hierarchical attention feature pyramid network for object detection","volume":"23","author":"J Dang","year":"2023","journal-title":"Sensors (Basel)"},{"key":"pone.0359400.ref031","doi-asserted-by":"crossref","first-page":"107677","DOI":"10.1016\/j.engappai.2023.107677","article-title":"IDP-Net: Industrial defect perception network based on cross-layer semantic information guidance and context concentration enhancement","volume":"130","author":"G Li","year":"2024","journal-title":"Eng Appl Artif Intell"},{"key":"pone.0359400.ref032","doi-asserted-by":"crossref","first-page":"106390","DOI":"10.1016\/j.engappai.2023.106390","article-title":"IDD-Net: Industrial defect detection method based on deep learning","volume":"123","author":"Z Zhang","year":"2023","journal-title":"Eng Appl Artif Intell"},{"issue":"9","key":"pone.0359400.ref033","doi-asserted-by":"crossref","first-page":"4267","DOI":"10.1007\/s00371-022-02589-w","article-title":"Multi-level feature fusion pyramid network for object detection","volume":"39","author":"Z Guo","year":"2023","journal-title":"Vis Comput"},{"issue":"11","key":"pone.0359400.ref034","doi-asserted-by":"crossref","first-page":"5114","DOI":"10.3390\/s23115114","article-title":"Mixed receptive fields augmented YOLO with multi-path spatial pyramid pooling for steel surface defect detection","volume":"23","author":"K Xia","year":"2023","journal-title":"Sensors (Basel)"},{"issue":"1","key":"pone.0359400.ref035","doi-asserted-by":"crossref","first-page":"7178","DOI":"10.1038\/s41598-023-34277-7","article-title":"Multi-scale semantic enhancement network for object detection","volume":"13","author":"D Guo","year":"2023","journal-title":"Sci Rep"},{"issue":"2","key":"pone.0359400.ref036","article-title":"FCPFNet: Feature complementation network with pyramid fusion for semantic segmentation","volume":"56","author":"J Lei","year":"2024","journal-title":"Neural Process Lett"},{"key":"pone.0359400.ref037","doi-asserted-by":"crossref","first-page":"109558","DOI":"10.1016\/j.patcog.2023.109558","article-title":"CANet: Contextual information and spatial attention based network for detecting small defects in manufacturing industry","volume":"140","author":"X Hou","year":"2023","journal-title":"Pattern Recognition"},{"key":"pone.0359400.ref038","doi-asserted-by":"crossref","first-page":"141764","DOI":"10.1109\/ACCESS.2023.3343566","article-title":"PDDD-Net: Defect detection network based on parallel attention mechanism and dual-channel spatial pyramid pooling","volume":"11","author":"T Sui","year":"2023","journal-title":"IEEE Access"},{"key":"pone.0359400.ref039","doi-asserted-by":"crossref","first-page":"88960","DOI":"10.1109\/ACCESS.2023.3303897","article-title":"Surface defect detection with channel-spatial attention modules and bi-directional feature pyramid","volume":"11","author":"H Xin","year":"2023","journal-title":"IEEE Access"},{"issue":"5","key":"pone.0359400.ref040","article-title":"Region\u2010based fully convolutional networks with deformable convolution and attention fusion for steel surface defect detection in industrial Internet of Things","volume":"17","author":"M Fu","year":"2023","journal-title":"IET Signal Processing"},{"issue":"10","key":"pone.0359400.ref041","doi-asserted-by":"crossref","first-page":"12250","DOI":"10.1109\/TPAMI.2023.3277881","article-title":"Prototype completion for few-shot learning","volume":"45","author":"B Zhang","year":"2023","journal-title":"IEEE Trans Pattern Anal Mach Intell"},{"key":"pone.0359400.ref042","doi-asserted-by":"crossref","first-page":"104757","DOI":"10.1016\/j.imavis.2023.104757","article-title":"WPE: Weighted prototype estimation for few-shot learning","volume":"137","author":"J Cao","year":"2023","journal-title":"Image and Vision Computing"},{"key":"pone.0359400.ref043","article-title":"Hierarchical adaptive correlation network for few-shot metal surface defect segmentation with cross-domain generalization","author":"K Gao","year":"2026","journal-title":"J Intell Manuf"},{"key":"pone.0359400.ref044","article-title":"Ultralytics YOLOv5","author":"G Jocher","year":"2020","journal-title":"Zenodo"},{"key":"pone.0359400.ref045","first-page":"91","article-title":"Faster R-CNN: Towards real-time object detection with region proposal networks","volume":"28","author":"S Ren","year":"2015","journal-title":"Adv Neural Inf Process Syst"},{"key":"pone.0359400.ref046","first-page":"1135","article-title":"Learning both weights and connections for efficient neural network","volume":"28","author":"S Han","year":"2015","journal-title":"Adv Neural Inf Process Syst"},{"key":"pone.0359400.ref047","article-title":"Distilling the knowledge in a neural network","author":"G Hinton","year":"2015"},{"key":"pone.0359400.ref048","doi-asserted-by":"crossref","unstructured":"Howard A, Sandler M, Chen B, Wang W, Chen L-C, Tan M, et al. Searching for MobileNetV3. In: 2019 IEEE\/CVF International Conference on Computer Vision (ICCV), 2019. 1314\u201324. https:\/\/doi.org\/10.1109\/iccv.2019.00140","DOI":"10.1109\/ICCV.2019.00140"},{"key":"pone.0359400.ref049","first-page":"232","article-title":"Infinite mixture prototypes for few-shot learning","volume":"97","author":"K Allen","year":"2019","journal-title":"Proceedings of the Machine Learning Research"},{"key":"pone.0359400.ref050","doi-asserted-by":"crossref","first-page":"29","DOI":"10.1007\/s44196-024-00423-w","article-title":"Deep learning-based integrated circuit surface defect detection: Addressing information density imbalance for industrial application","volume":"17","author":"X Wang","year":"2024","journal-title":"Int J Comput Intell Syst"},{"issue":"10","key":"pone.0359400.ref051","doi-asserted-by":"crossref","first-page":"7742","DOI":"10.1109\/TII.2025.3574815","article-title":"Few-shot parameter efficient finetuning for SAM in salient steel surface defect detection","volume":"21","author":"J Su","year":"2025","journal-title":"IEEE Trans Ind Inf"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0359400","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T17:29:12Z","timestamp":1790616552000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0359400"}},"subtitle":[],"editor":[{"given":"Xiaowei","family":"Li","sequence":"first","affiliation":[{"name":"Sichuan University","place":["CHINA"]}],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,9,28]]},"references-count":51,"journal-issue":{"issue":"9","published-online":{"date-parts":[[2026,9,28]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0359400","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,28]]}}}