{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,9]],"date-time":"2026-01-09T07:59:14Z","timestamp":1767945554128,"version":"3.49.0"},"reference-count":34,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,1,9]],"date-time":"2026-01-09T00:00:00Z","timestamp":1767916800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Railway transportation is increasingly critical for modern urban and intercity mobility. However, the expanding scale and intensifying operational intensity of rail networks have elevated track defect detection to a key concern. Traditional inspection methods (manual, ultrasonic, eddy current, magnetic flux leakage testing) are limited by insufficient accuracy, low efficiency, or poor adaptability to complex environmental conditions.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>An enhanced defect detection framework based on an improved YOLOv8 algorithm was proposed, tailored for small targets and complex backgrounds. Three core improvements were integrated: 1) AVCStem module with variable convolution kernels to dynamically adapt to defects of different shapes and scales; 2) ADSPPF module using multi-scale pooling and multi-branch attention mechanisms to preserve fine-grained features across scales; 3) MSF module for enhanced multi-scale feature fusion via partial convolution and hierarchical feature alignment.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results and Discussion<\/jats:title>\n                    <jats:p>Experiments on a real-world track defect dataset showed the proposed model achieved 90.2% detection precision, 90.2% mAP@0.5, and 73.2% mAP@0.5:0.95. Meanwhile, the model size was reduced to 5.2MB with 2.45M parameters. Comparative and ablation studies confirmed the complementary advantages of each module and the model\u2019s superior performance over existing lightweight detectors. The proposed model provides a robust, accurate, and efficient solution for real-time railway defect detection. It exhibits strong potential for deployment in edge AI devices and mobile inspection robots, addressing the limitations of traditional inspection methods.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2025.1711309","type":"journal-article","created":{"date-parts":[[2026,1,9]],"date-time":"2026-01-09T06:36:14Z","timestamp":1767940574000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["An improved YOLOv8n with multi-scale feature fusion for real time and high precision railway track defect detection"],"prefix":"10.3389","volume":"8","author":[{"given":"Zhihong","family":"Zhang","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Liling","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xin","family":"Lu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tingting","family":"Ma","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Feng","family":"Huang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sheng","family":"Zhong","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2026,1,9]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"105965","DOI":"10.1016\/j.engappai.2023.105965","article-title":"MSRConvNet: classification of railway track defects using multi-scale residual convolutional neural network","volume":"121","author":"Acikgoz","year":"2023","journal-title":"Eng. Appl. Artif. Intell."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"7937","DOI":"10.3390\/s21237937","article-title":"Detection and classification system for rail surface defects based on Eddy current","volume":"21","author":"Alvarenga","year":"2021","journal-title":"Sensors"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"107706","DOI":"10.1016\/j.asoc.2021.107706","article-title":"Defect classification based on deep features for railway tracks in sustainable transportation","volume":"111","author":"Aydin","year":"2021","journal-title":"Appl. Soft Comput."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"1437","DOI":"10.3390\/e23111437","article-title":"A study on railway surface defects detection based on machine vision","volume":"23","author":"Bai","year":"2021","journal-title":"Entropy"},{"key":"ref5","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR52729.2023.01157","article-title":"Run, Don't walk: chasing higher FLOPS for faster neural networks","author":"Chen","year":"2023"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"3916","DOI":"10.3390\/s23083916","article-title":"Recent advances in wayside railway wheel flat detection techniques: a review","volume":"23","author":"Fu","year":"2023","journal-title":"Sensors"},{"key":"ref7","doi-asserted-by":"publisher","first-page":"153","DOI":"10.1109\/TITS.2016.2568758","article-title":"Deep multitask learning for railway track inspection","volume":"18","author":"Gibert","year":"2016","journal-title":"IEEE Trans. Intell. Transp. Syst."},{"key":"ref8","doi-asserted-by":"publisher","first-page":"137","DOI":"10.1016\/j.measurement.2017.08.040","article-title":"Digital image analysis technique for measuring railway track defects and ballast gradation","volume":"113","author":"Guerrieri","year":"2018","journal-title":"Measurement"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"1367","DOI":"10.3390\/s20051367","article-title":"A rail fastener tightness detection approach using multi-source visual sensor","volume":"20","author":"Han","year":"2020","journal-title":"Sensors"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"103326","DOI":"10.1016\/j.rineng.2024.103326","article-title":"An end-to-end approach to detect railway track defects based on supervised and self-supervised learning","volume":"24","author":"Haroon","year":"2024","journal-title":"Results Eng."},{"key":"ref11","doi-asserted-by":"publisher","first-page":"8796","DOI":"10.1109\/TITS.2024.3368213","article-title":"PCSGAN: a perceptual constrained generative model for railway defect sample expansion from a single image","volume":"25","author":"He","year":"2024","journal-title":"IEEE Trans. Intell. Transp. Syst."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"105057","DOI":"10.1016\/j.imavis.2024.105057","article-title":"ASF-YOLO: a novel YOLO model with attentional scale sequence fusion for cell instance segmentation","volume":"147","author":"Kang","year":"2024","journal-title":"Image Vis. Comput."},{"key":"ref13","doi-asserted-by":"publisher","first-page":"167","DOI":"10.12700\/APH.19.3.2022.3.14","article-title":"A review of research on detection and evaluation of the rail surface defects","volume":"19","author":"Kou","year":"2022","journal-title":"Acta Polytech. Hung."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"207","DOI":"10.1016\/j.ijtst.2024.06.006","article-title":"A systematic literature review of defect detection in railways using machine vision-based inspection methods","volume":"18","author":"Kumar","year":"2025","journal-title":"Int. J. Transp. Sci. Technol."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"62","DOI":"10.1007\/s11554-024-01436-6","article-title":"Slim-neck by GSConv: a lightweight-design for real-time detector architectures","volume":"21","author":"Li","year":"2022","journal-title":"J. Real Time Image Proc."},{"key":"ref16","doi-asserted-by":"publisher","first-page":"7894","DOI":"10.3390\/s23187894","article-title":"Fs-rsdd: few-shot rail surface defect detection with prototype learning","volume":"23","author":"Min","year":"2023","journal-title":"Sensors"},{"key":"ref17","article-title":"SENetV2: aggregated dense layer for channelwise and global representations","author":"Narayanan","year":"2023"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"1694","DOI":"10.1109\/TII.2021.3085848","article-title":"Attention network for rail surface defect detection via consistency of intersection-over-Union(IoU)-guided center-point estimation","volume":"18","author":"Ni","year":"2021","journal-title":"IEEE Trans. Ind. Inform."},{"key":"ref19","doi-asserted-by":"publisher","first-page":"192","DOI":"10.3390\/jimaging10080192","article-title":"RailTrack-DaViT: a vision transformer-based approach for automated railway track defect detection","volume":"10","author":"Phaphuangwittayakul","year":"2024","journal-title":"J. Imag."},{"key":"ref20","doi-asserted-by":"publisher","first-page":"110983","DOI":"10.1016\/j.ymssp.2023.110983","article-title":"A review of distributed acoustic sensing applications for railroad condition monitoring","volume":"208","author":"Rahman","year":"2024","journal-title":"Mech. Syst. Signal Process."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"4177","DOI":"10.3390\/app10124177","article-title":"Small-scale face detection based on improved R-FCN","volume":"10","author":"Tang","year":"2020","journal-title":"Appl. Sci."},{"key":"ref22","doi-asserted-by":"publisher","first-page":"113091","DOI":"10.1016\/j.sna.2021.113091","article-title":"A novel magnetic flux leakage method based on the ferromagnetic lift-off layer with through groove","volume":"332","author":"Tang","year":"2021","journal-title":"Sens. Actuators A Phys."},{"key":"ref23","doi-asserted-by":"publisher","first-page":"1680","DOI":"10.3390\/make5040083","article-title":"A comprehensive review of YOLO architectures in computer vision: from YOLOv1 to YOLOv8 and YOLO-NAS","volume":"5","author":"Terven","year":"2023","journal-title":"Mach. Learn. Knowl. Extr."},{"key":"ref24","doi-asserted-by":"publisher","first-page":"4874","DOI":"10.1109\/TMECH.2022.3167412","article-title":"Collaborative learning attention network based on RGB image and depth image for surface defect inspection of no-service rail","volume":"27","author":"Wang","year":"2022","journal-title":"IEEE\/ASME Trans. Mechatron."},{"key":"ref25","doi-asserted-by":"publisher","first-page":"44984","DOI":"10.1109\/ACCESS.2024.3380009","article-title":"An improved YOLOv8 algorithm for rail surface defect detection","volume":"12","author":"Wang","year":"2024","journal-title":"IEEE Access"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"126451","DOI":"10.1109\/ACCESS.2022.3224594","article-title":"Detection of surface defects on railway tracks based on deep learning","volume":"10","author":"Wang","year":"2022","journal-title":"IEEE Access"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"3514909","DOI":"10.1109\/TIM.2023.3269125","article-title":"Laser-induced ultrasonic guided waves based corrosion diagnosis of rail foot","volume":"72","author":"Wang","year":"2023","journal-title":"IEEE Trans. Instrum. Meas."},{"key":"ref28","doi-asserted-by":"publisher","first-page":"1169","DOI":"10.1109\/TIP.2020.3042065","article-title":"CGnet: a light-weight context guided network for semantic segmentation","volume":"30","author":"Wu","year":"2020","journal-title":"IEEE Trans. Image Process."},{"key":"ref29","doi-asserted-by":"publisher","first-page":"4627","DOI":"10.3390\/s23104627","article-title":"Detection of rail defects using NDT methods","volume":"23","author":"Xiong","year":"2023","journal-title":"Sensors"},{"key":"ref30","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TIM.2021.3138498","article-title":"Deep learning and machine vision-based inspection of rail surface defects","volume":"71","author":"Yang","year":"2022","journal-title":"IEEE Trans. Instrum. Meas."},{"key":"ref31","doi-asserted-by":"publisher","first-page":"21157","DOI":"10.1109\/JSEN.2024.3402730","article-title":"Research on real-time detection system of rail surface defects based on deep learning","volume":"24","author":"Yaodong","year":"2024","journal-title":"IEEE Sensors J."},{"key":"ref32","doi-asserted-by":"publisher","first-page":"2200027","DOI":"10.1002\/adts.202200027","article-title":"Railway track vibration analysis and intelligent recognition of fastener defects","volume":"5","author":"Yin","year":"2022","journal-title":"Adv. Theory Simul."},{"key":"ref33","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TIM.2024.3373083","article-title":"A tightness detection method for railway fasteners based on RGB-P bimodal data","volume":"73","author":"Yuan","year":"2024","journal-title":"IEEE Trans. Instrum. Meas."},{"key":"ref34","doi-asserted-by":"publisher","first-page":"2564","DOI":"10.3390\/electronics13132564","article-title":"Rail surface defect detection based on dual-path feature fusion","volume":"13","author":"Zhong","year":"2024","journal-title":"Electronics"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1711309\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,9]],"date-time":"2026-01-09T06:36:15Z","timestamp":1767940575000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1711309\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,9]]},"references-count":34,"alternative-id":["10.3389\/frai.2025.1711309"],"URL":"https:\/\/doi.org\/10.3389\/frai.2025.1711309","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,9]]},"article-number":"1711309"}}