Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR

Abstract Filters are critical components in automotive engine systems, responsible for maintaining stable operation by removing impurities from liquids and gases. Their performance is highly sensitive to surface defects, rendering high-precision automated inspection essential. However, existing defe...

Full description

Saved in:
Bibliographic Details
Main Authors: Maoyuan Zhang, Xiaojuan Wei, Guojun Liu, Mengxu Chen, Chunxia Zhao, Yingxiao Liu, Zhikang Bao, Yunfeng Guo, Run An, Pengcheng Zhao
Format: Article
Language:English
Published: Nature Portfolio 2025-08-01
Series:Scientific Reports
Online Access:https://doi.org/10.1038/s41598-025-13960-x
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849344846489190400
author Maoyuan Zhang
Xiaojuan Wei
Guojun Liu
Mengxu Chen
Chunxia Zhao
Yingxiao Liu
Zhikang Bao
Yunfeng Guo
Run An
Pengcheng Zhao
author_facet Maoyuan Zhang
Xiaojuan Wei
Guojun Liu
Mengxu Chen
Chunxia Zhao
Yingxiao Liu
Zhikang Bao
Yunfeng Guo
Run An
Pengcheng Zhao
author_sort Maoyuan Zhang
collection DOAJ
description Abstract Filters are critical components in automotive engine systems, responsible for maintaining stable operation by removing impurities from liquids and gases. Their performance is highly sensitive to surface defects, rendering high-precision automated inspection essential. However, existing defect detection algorithms often struggle to balance between detection accuracy and the computational efficiency required for industrial deployment. To address this trade-off, this study introduces an improved detection method based on the Real-Time DEtection TRansformer(RT-DETR) framework. First, a large-kernel attention mechanism is integrated into the backbone to enhance multi-scale feature extraction and fusion, while reducing architectural redundancy. Second, the RepC3 structure within the cross-scale fusion module is replaced with a module based on the generalized-efficient layer aggregation network that uses a more efficient layer aggregation strategy to improve feature localization. Finally, the Adown downsampling module is introduced, employing a multi-path design that reduces parameter count while preserving critical feature details during scale reduction. Experimental results on our industrial filter surface defect dataset show that the enhanced RT-DETR model achieves a mean average precision of 97.6%, a 7.3 percentage point increase over the baseline. Furthermore, the model reduces parameter count by 6.9% and computational load by 13.1%, demonstrating its improved efficiency. Generalization experiments on the public NEU-DET dataset and GC10-DET dataset further confirm the model’s robustness and effectiveness, demonstrating its suitability for industrial applications requiring both high accuracy and lightweight deployment.
format Article
id doaj-art-2e4c88e1855e490aba90bd42e7da9b70
institution Kabale University
issn 2045-2322
language English
publishDate 2025-08-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-2e4c88e1855e490aba90bd42e7da9b702025-08-20T03:42:34ZengNature PortfolioScientific Reports2045-23222025-08-0115112110.1038/s41598-025-13960-xBalancing complexity and accuracy for defect detection on filters with an improved RT-DETRMaoyuan Zhang0Xiaojuan Wei1Guojun Liu2Mengxu Chen3Chunxia Zhao4Yingxiao Liu5Zhikang Bao6Yunfeng Guo7Run An8Pengcheng Zhao9College of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityZhejiang Zhenhang Industrial Group Company Ltd.Zhejiang Zhenhang Industrial Group Company Ltd.College of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityCollege of Electrical Engineering, Northwest Minzu UniversityAbstract Filters are critical components in automotive engine systems, responsible for maintaining stable operation by removing impurities from liquids and gases. Their performance is highly sensitive to surface defects, rendering high-precision automated inspection essential. However, existing defect detection algorithms often struggle to balance between detection accuracy and the computational efficiency required for industrial deployment. To address this trade-off, this study introduces an improved detection method based on the Real-Time DEtection TRansformer(RT-DETR) framework. First, a large-kernel attention mechanism is integrated into the backbone to enhance multi-scale feature extraction and fusion, while reducing architectural redundancy. Second, the RepC3 structure within the cross-scale fusion module is replaced with a module based on the generalized-efficient layer aggregation network that uses a more efficient layer aggregation strategy to improve feature localization. Finally, the Adown downsampling module is introduced, employing a multi-path design that reduces parameter count while preserving critical feature details during scale reduction. Experimental results on our industrial filter surface defect dataset show that the enhanced RT-DETR model achieves a mean average precision of 97.6%, a 7.3 percentage point increase over the baseline. Furthermore, the model reduces parameter count by 6.9% and computational load by 13.1%, demonstrating its improved efficiency. Generalization experiments on the public NEU-DET dataset and GC10-DET dataset further confirm the model’s robustness and effectiveness, demonstrating its suitability for industrial applications requiring both high accuracy and lightweight deployment.https://doi.org/10.1038/s41598-025-13960-x
spellingShingle Maoyuan Zhang
Xiaojuan Wei
Guojun Liu
Mengxu Chen
Chunxia Zhao
Yingxiao Liu
Zhikang Bao
Yunfeng Guo
Run An
Pengcheng Zhao
Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
Scientific Reports
title Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
title_full Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
title_fullStr Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
title_full_unstemmed Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
title_short Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
title_sort balancing complexity and accuracy for defect detection on filters with an improved rt detr
url https://doi.org/10.1038/s41598-025-13960-x
work_keys_str_mv AT maoyuanzhang balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT xiaojuanwei balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT guojunliu balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT mengxuchen balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT chunxiazhao balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT yingxiaoliu balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT zhikangbao balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT yunfengguo balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT runan balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr
AT pengchengzhao balancingcomplexityandaccuracyfordefectdetectiononfilterswithanimprovedrtdetr