{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T22:58:54Z","timestamp":1788389934794,"version":"build-2803163510"},"reference-count":34,"publisher":"PeerJ","license":[{"start":{"date-parts":[[2026,1,27]],"date-time":"2026-01-27T00:00:00Z","timestamp":1769472000000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"College Students\u2019 Innovative Entrepreneurial Training Plan Program","award":["gzusc2025052"],"award-info":[{"award-number":["gzusc2025052"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"crossref","award":["62373116"],"award-info":[{"award-number":["62373116"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"crossref"}]},{"name":"Guizhou Provincial Science and Technology Projects, China","award":["QKHZC [2023]118, PTRC [2020]6007-2"],"award-info":[{"award-number":["QKHZC [2023]118, PTRC [2020]6007-2"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>\n                    To address the primary challenges of inadequate multi-scale defect perception capabilities and limited computational resources on edge devices in steel surface defect detection scenarios, this article proposes a steel surface defect detection based on multi-scale dynamic convolution and lightweight cross-stage fusion (LightSDD). First, to adaptively enhance the feature representation ability of small targets, the multi-scale dynamic convolution kernel is fused with the channel attention mechanism, and a fine-grained feature fusion method based on C3k2\u00a0multi-scale dynamic convolution is proposed. Next, to reduce redundant parameters while retaining key positioning information, an interpolation alignment strategy is adopted for multi-resolution features, along with channel compression and cross-resolution feature concatenation, based on the replacement of the original you only look once version 11 (YOLOV11) detection head. This approach enables the design of a lightweight cross-stage detection head, LSCD. Later, comparative experiments on the NEU-DET and GC10-DET datasets demonstrate that LightSDD outperforms seven state-of-the-art methods. Finally, performance tests on edge devices indicate that LightSDD exhibits strong robustness and practicality. Code is available at:\n                    <jats:uri xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"https:\/\/github.com\/chaoszzz-zyx\/LightSDD\">https:\/\/github.com\/chaoszzz-zyx\/LightSDD<\/jats:uri>\n                    (DOI:\n                    <jats:uri xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"http:\/\/dx.doi.org\/10.5281\/zenodo.17012588\">10.5281\/zenodo.17012588<\/jats:uri>\n                    ).\n                  <\/jats:p>","DOI":"10.7717\/peerj-cs.3485","type":"journal-article","created":{"date-parts":[[2026,1,27]],"date-time":"2026-01-27T08:15:21Z","timestamp":1769501721000},"page":"e3485","source":"Crossref","is-referenced-by-count":1,"title":["Steel surface defect detection based on multi-scale dynamic convolution and lightweight cross-stage fusion"],"prefix":"10.7717","volume":"12","author":[{"given":"Yuxuan","family":"Zhang","sequence":"first","affiliation":[{"name":"School of Mechanical Engineering, Guizhou University, Guiyang, Guizhou, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Biao","family":"Xu","sequence":"additional","affiliation":[{"name":"Key Laboratory of Advanced Manufacturing Technology of the Ministry of Education, Guizhou University, Guiyang, Guizhou, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8761-5195","authenticated-orcid":true,"given":"Guanci","family":"Yang","sequence":"additional","affiliation":[{"name":"Key Laboratory of Advanced Manufacturing Technology of the Ministry of Education, Guizhou University, Guiyang, Guizhou, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"4443","published-online":{"date-parts":[[2026,1,27]]},"reference":[{"key":"10.7717\/peerj-cs.3485\/ref-1","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/AVSS.2019.8909834","article-title":"High efficient single-stage steel surface defect detection","author":"Akhyar","year":"2019"},{"issue":"7553","key":"10.7717\/peerj-cs.3485\/ref-2","doi-asserted-by":"publisher","first-page":"112","DOI":"10.1016\/j.neucom.2019.10.008","article-title":"Impact of fully connected layers on performance of convolutional neural networks for image classification","volume":"378","author":"Basha","year":"2020","journal-title":"Neurocomputing"},{"issue":"2","key":"10.7717\/peerj-cs.3485\/ref-3","doi-asserted-by":"publisher","first-page":"110795","DOI":"10.1016\/j.enbuild.2021.110795","article-title":"Fault detection and diagnosis for Air Handling Unit based on multiscale convolutional neural networks","volume":"236","author":"Cheng","year":"2021","journal-title":"Energy and Buildings"},{"issue":"5","key":"10.7717\/peerj-cs.3485\/ref-4","doi-asserted-by":"publisher","first-page":"1874","DOI":"10.3390\/app14051874","article-title":"Deep learning (Fast R-CNN)-based evaluation of rail surface defects","volume":"14","author":"Choi","year":"2024","journal-title":"Applied Sciences"},{"issue":"16","key":"10.7717\/peerj-cs.3485\/ref-5","doi-asserted-by":"publisher","first-page":"128018","DOI":"10.1016\/j.eswa.2025.128018","article-title":"MirageNet: improving the network performance of image understanding with very low FLOPs","volume":"286","author":"Fu","year":"2025","journal-title":"Expert Systems with Applications"},{"issue":"16","key":"10.7717\/peerj-cs.3485\/ref-6","doi-asserted-by":"publisher","first-page":"6941","DOI":"10.3390\/app14166941","article-title":"Lightweight frequency recalibration network for diabetic retinopathy multi-lesion segmentation","volume":"14","author":"Fu","year":"2024","journal-title":"Applied Sciences"},{"key":"10.7717\/peerj-cs.3485\/ref-7","first-page":"580","article-title":"Rich feature hierarchies for accurate object detection and semantic segmentation","author":"Girshick","year":"2014"},{"issue":"4","key":"10.7717\/peerj-cs.3485\/ref-8","doi-asserted-by":"publisher","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":"He","year":"2019","journal-title":"IEEE Transactions on Instrumentation and Measurement"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-9","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1016\/j.sysarc.2019.02.008","article-title":"Filter-based deep-compression with global average pooling for convolutional networks","volume":"95","author":"Hsiao","year":"2019","journal-title":"Journal of Systems Architecture"},{"issue":"10","key":"10.7717\/peerj-cs.3485\/ref-10","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TIM.2024.3488136","article-title":"SSA-YOLO: an improved YOLO for hot-rolled strip steel surface defect detection","volume":"73","author":"Huang","year":"2024","journal-title":"IEEE Transactions on Instrumentation and Measurement"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-11","doi-asserted-by":"publisher","first-page":"108159","DOI":"10.1016\/j.patcog.2021.108159","article-title":"Delving deep into spatial pooling for squeeze-and-excitation networks","volume":"121","author":"Jin","year":"2022","journal-title":"Pattern Recognition"},{"key":"10.7717\/peerj-cs.3485\/ref-12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2410.17725","article-title":"Yolov11: an overview of the key architectural enhancements","author":"Khanam","year":"2024"},{"issue":"1","key":"10.7717\/peerj-cs.3485\/ref-13","doi-asserted-by":"publisher","first-page":"110979","DOI":"10.1016\/j.patcog.2024.110979","article-title":"Global attention module and cascade fusion network for steel surface defect detection","volume":"158","author":"Liu","year":"2025","journal-title":"Pattern Recognition"},{"key":"10.7717\/peerj-cs.3485\/ref-14","first-page":"209","article-title":"Surface defect detection method of hot rolling strip based on improved SSD model","author":"Liu","year":"2021"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-15","doi-asserted-by":"publisher","first-page":"115060","DOI":"10.1016\/j.measurement.2024.115060","article-title":"WSS-YOLO: an improved industrial defect detection network for steel surface defects","volume":"236","author":"Lu","year":"2024","journal-title":"Measurement"},{"issue":"4","key":"10.7717\/peerj-cs.3485\/ref-16","doi-asserted-by":"publisher","first-page":"109942","DOI":"10.1016\/j.engappai.2024.109942","article-title":"Intra-and inter-instance Location Correlation Network for human-object interaction detection","volume":"142","author":"Lu","year":"2025","journal-title":"Engineering Applications of Artificial Intelligence"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-17","doi-asserted-by":"publisher","first-page":"117608","DOI":"10.1016\/j.measurement.2025.117608","article-title":"SFW-YOLO: a lightweight multi-scale dynamic attention network for weld defect detection in steel bridge inspection","volume":"253","author":"Luo","year":"2025","journal-title":"Measurement"},{"issue":"2","key":"10.7717\/peerj-cs.3485\/ref-18","doi-asserted-by":"publisher","first-page":"103377","DOI":"10.1016\/j.aei.2025.103377","article-title":"ELA-YOLO: an efficient method with linear attention for steel surface defect detection during manufacturing","volume":"65","author":"Ma","year":"2025","journal-title":"Advanced Engineering Informatics"},{"key":"10.7717\/peerj-cs.3485\/ref-19","doi-asserted-by":"publisher","first-page":"504","DOI":"10.1016\/j.asoc.2014.11.023","article-title":"A systematic review of machine learning techniques for software fault prediction","volume":"27","author":"Malhotra","year":"2015","journal-title":"Applied Soft Computing"},{"issue":"2","key":"10.7717\/peerj-cs.3485\/ref-20","doi-asserted-by":"publisher","first-page":"102325","DOI":"10.1016\/j.displa.2022.102325","article-title":"The quantitative research on length and area perception: a guidance on shape encoding in visual interface","volume":"75","author":"Qi","year":"2022","journal-title":"Displays"},{"issue":"7","key":"10.7717\/peerj-cs.3485\/ref-21","doi-asserted-by":"publisher","first-page":"926","DOI":"10.1109\/LSP.2018.2822810","article-title":"Additive margin softmax for face verification","volume":"25","author":"Wang","year":"2018","journal-title":"IEEE Signal Processing Letters"},{"issue":"2","key":"10.7717\/peerj-cs.3485\/ref-22","doi-asserted-by":"publisher","first-page":"104","DOI":"10.1016\/j.neunet.2023.11.065","article-title":"Towards performance-maximizing neural network pruning via global channel attention","volume":"171","author":"Wang","year":"2024","journal-title":"Neural Networks"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-23","doi-asserted-by":"publisher","first-page":"108269","DOI":"10.1016\/j.compeleceng.2022.108269","article-title":"Rail surface defect detection based on improved Mask R-CNN","volume":"102","author":"Wang","year":"2022","journal-title":"Computers and Electrical Engineering"},{"issue":"4","key":"10.7717\/peerj-cs.3485\/ref-24","doi-asserted-by":"publisher","first-page":"041002","DOI":"10.1115\/1.4063860","article-title":"Metal surface defect detection method based on improved cascade R-CNN","volume":"24","author":"Wang","year":"2024","journal-title":"Journal of Computing and Information Science in Engineering"},{"issue":"11","key":"10.7717\/peerj-cs.3485\/ref-25","doi-asserted-by":"publisher","first-page":"109886","DOI":"10.1016\/j.sigpro.2025.109886","article-title":"TDU-DLNet: a transformer-based deep unfolding network for dictionary learning","volume":"231","author":"Wu","year":"2025","journal-title":"Signal Processing"},{"issue":"3","key":"10.7717\/peerj-cs.3485\/ref-26","doi-asserted-by":"publisher","first-page":"128775","DOI":"10.1016\/j.neucom.2024.128775","article-title":"An efficient re-parameterization feature pyramid network on YOLOv8 to the detection of steel surface defect","volume":"614","author":"Xie","year":"2025","journal-title":"Neurocomputing"},{"issue":"5","key":"10.7717\/peerj-cs.3485\/ref-27","doi-asserted-by":"publisher","first-page":"055017","DOI":"10.1088\/1361-6501\/ad296d","article-title":"A light weight multi-scale feature fusion steel surface defect detection model based on YOLOv8","volume":"35","author":"Xie","year":"2024","journal-title":"Measurement Science and Technology"},{"issue":"2","key":"10.7717\/peerj-cs.3485\/ref-28","doi-asserted-by":"publisher","first-page":"683","DOI":"10.1109\/TAFFC.2024.3458464","article-title":"Multiscale facial expression recognition based on dynamic global and static local attention","volume":"16","author":"Xu","year":"2025","journal-title":"IEEE Transactions on Affective Computing"},{"issue":"11","key":"10.7717\/peerj-cs.3485\/ref-29","doi-asserted-by":"publisher","first-page":"102721","DOI":"10.1016\/j.inffus.2024.102721","article-title":"Interpretability research of deep learning: a literature survey","volume":"115","author":"Xu","year":"2025","journal-title":"Information Fusion"},{"key":"10.7717\/peerj-cs.3485\/ref-30","first-page":"353","article-title":"Steel surface defects detection based on improved faster R-CNN","author":"Xu","year":"2022"},{"issue":"24","key":"10.7717\/peerj-cs.3485\/ref-31","doi-asserted-by":"publisher","first-page":"5755","DOI":"10.3390\/ma13245755","article-title":"Using deep learning to detect defects in manufacturing: a comprehensive survey and current challenges","volume":"13","author":"Yang","year":"2020","journal-title":"Materials"},{"key":"10.7717\/peerj-cs.3485\/ref-32","doi-asserted-by":"publisher","first-page":"97","DOI":"10.1016\/j.neucom.2022.07.051","article-title":"Channel pruning based on convolutional neural network sensitivity","volume":"507","author":"Yang","year":"2022","journal-title":"Neurocomputing"},{"issue":"4","key":"10.7717\/peerj-cs.3485\/ref-33","doi-asserted-by":"publisher","first-page":"109289","DOI":"10.1016\/j.engappai.2024.109289","article-title":"A novel real-time steel surface defect detection method with enhanced feature extraction and adaptive fusion","volume":"138","author":"Yang","year":"2024","journal-title":"Engineering Applications of Artificial Intelligence"},{"issue":"4","key":"10.7717\/peerj-cs.3485\/ref-34","doi-asserted-by":"publisher","first-page":"115189","DOI":"10.1016\/j.measurement.2024.115189","article-title":"Steel surface defect detection algorithm in complex background scenarios","volume":"237","author":"Zhao","year":"2024","journal-title":"Measurement"}],"container-title":["PeerJ Computer Science"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/peerj.com\/articles\/cs-3485.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3485.xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3485.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3485.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,27]],"date-time":"2026-01-27T08:15:26Z","timestamp":1769501726000},"score":1,"resource":{"primary":{"URL":"https:\/\/peerj.com\/articles\/cs-3485"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,27]]},"references-count":34,"alternative-id":["10.7717\/peerj-cs.3485"],"URL":"https:\/\/doi.org\/10.7717\/peerj-cs.3485","archive":["CLOCKSS","LOCKSS","Portico"],"relation":{},"ISSN":["2376-5992"],"issn-type":[{"value":"2376-5992","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,27]]},"article-number":"e3485"}}