A weakly supervised learning pipeline for profiled fibre inspection

被引:1
作者
Chen, Zhao [1 ,2 ]
Xiu, Yahui [1 ]
Zheng, Yuxin [1 ]
Wang, Xinxin [1 ]
Wang, Qian [1 ,3 ]
Guo, Danqi [1 ]
Wan, Yan [1 ]
机构
[1] Donghua Univ, Sch Comp Sci & Technol, Shanghai 201620, Peoples R China
[2] Univ Warwick, Dept Comp Sci, Coventry, England
[3] East China Normal Univ, Shanghai Key Lab Multidimens Informat Proc, Shanghai, Peoples R China
基金
中央高校基本科研业务费专项资金资助;
关键词
composite similarity measurement; profiled fibre recognition; shape factor estimation; weakly supervised learning; CROSS-SECTION; SHAPE; IDENTIFICATION; SEGMENTATION;
D O I
10.1049/ipr2.12984
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Automatic profiled fibre recognition and analysis can accelerate quality inspection and contributes to the upgrade of the textile industry. However, these tasks often require significant manual effort to generate instance-level annotations for fully supervised training. In this paper, the authors propose a weakly supervised pipeline for profiled fibre inspection using electron-microscopic (EM) images with only image-level annotations. It automatically identifies fibre instances and estimates shape factors to facilitate fibre quality inspection. As the core of the pipeline, the weakly supervised network (WesNet) is designed to localize hundreds of crowded fibre samples by raw patch generation and fibre sample sifting. Particularly, the composite similarity measurement integrates different patch-wise similarities, enabling the network to distinguish fibre from background robustly. For quality inspection, the pipeline further analyzes the fibre instances, utilizing several efficient techniques to estimate the shape factors. Experiments on the real fibre electron-microscopic images demonstrate the efficacy and efficiency of the pipeline. Results show that WesNet outperforms several supervised and weakly supervised methods, including two state-of-the-art weakly supervised networks. This paper proposes a novel weakly supervised pipeline for profiled fibre recognition and inspection. A weakly supervised network (WesNet) is designed to localize fibre instances in electron-microscopic images with only image-level annotations. The framework automatically measures shape factors to provide reference for fibre quality inspection.image
引用
收藏
页码:772 / 784
页数:13
相关论文
共 50 条
[1]  
Aguilar C., 2021, P IEEECVF C COMPUTER
[2]   Measuring the Objectness of Image Windows [J].
Alexe, Bogdan ;
Deselaers, Thomas ;
Ferrari, Vittorio .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2012, 34 (11) :2189-2202
[3]   Microscopy analysis neural network to solve detection, enumeration and segmentation from image-level annotations [J].
Bilodeau, Anthony ;
Delmas, Constantin V. L. ;
Parent, Martin ;
De Koninck, Paul ;
Durand, Audrey ;
Lavoie-Cardinal, Flavie .
NATURE MACHINE INTELLIGENCE, 2022, 4 (05) :455-+
[4]   Influence of the shape of fiber cross section on fabric surface characteristics [J].
Bueno, MA ;
Aneja, AP ;
Renner, M .
JOURNAL OF MATERIALS SCIENCE, 2004, 39 (02) :557-564
[6]   Weakly supervised vehicle detection in satellite images via multi-instance discriminative learning [J].
Cao, Liujuan ;
Luo, Feng ;
Chen, Li ;
Sheng, Yihan ;
Wang, Haibin ;
Wang, Cheng ;
Ji, Rongrong .
PATTERN RECOGNITION, 2017, 64 :417-424
[7]   Real-World Anomaly Detection by Using Digital Twin Systems and Weakly Supervised Learning [J].
Castellani, Andrea ;
Schmitt, Sebastian ;
Squartini, Stefano .
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, 2021, 17 (07) :4733-4742
[8]   Functional Fiber Materials to Smart Fiber Devices [J].
Chen, Chuanrui ;
Feng, Jianyou ;
Li, Jiaxin ;
Guo, Yue ;
Shi, Xiang ;
Peng, Huisheng .
CHEMICAL REVIEWS, 2023, 123 (02) :613-662
[9]  
Chen W., 2018, COTTON TEXTILE TECHN, V46, P13
[10]  
[陈晓春 Chen Xiaochun], 2017, [纺织学报, Journal of Textile Research], V38, P143