Recognition of Defective Mineral Wool Using Pruned ResNet Models

Mehdi Rafiei; Dat Thanh Tran; Alexandros Iosifidis

doi:10.1109/INDIN51400.2023.10217993

Recognition of Defective Mineral Wool Using Pruned ResNet Models

Mehdi Rafiei, Dat Thanh Tran, Alexandros Iosifidis

Computing Sciences

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

Abstract

Mineral wool production is a non-linear process that makes it hard to control the final quality. Therefore, having a nondestructive method to analyze the product quality and recognize defective products is critical. For this purpose, we developed a visual quality control system for mineral wool. X-ray images of wool specimens were collected to create a training set of defective and non-defective samples. Afterward, we developed several recognition models based on the ResNet architecture to find the most efficient model. In order to have a light-weight and fast inference model for real-life applicability, two structural pruning methods are applied to the classifiers. Considering the low quantity of the dataset, cross-validation and augmentation methods are used during the training. As a result, we obtained a model with more than 98% accuracy, which in comparison to the current procedure used at the company, it can recognize 20% more defective products.

Original language	English
Title of host publication	2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023
Editors	Helene Dorksen, Stefano Scanzio, Jurgen Jasperneite, Lukasz Wisniewski, Kim Fung Man, Thilo Sauter, Lucia Seno, Henning Trsek, Valeriy Vyatkin
Publisher	IEEE
ISBN (Electronic)	978-1-6654-9313-0
DOIs	https://doi.org/10.1109/INDIN51400.2023.10217993
Publication status	Published - 2023
Publication type	A4 Article in conference proceedings
Event	IEEE International Conference on Industrial Informatics - Lemgo, Germany Duration: 17 Jul 2023 → 20 Jul 2023

Publication series

Name	IEEE International Conference on Industrial Informatics (INDIN)
Volume	2023-July
ISSN (Print)	1935-4576

Conference

Conference	IEEE International Conference on Industrial Informatics
Country/Territory	Germany
City	Lemgo
Period	17/07/23 → 20/07/23

Funding

The research leading to the results of this paper received funding from the Innovation Fund Denmark as part of MADE FAST. We thank ROCKWOOL Group company for their help to provide required knowledge regarding the problem and their current methods, and also the support to provide product samples for the test. We thank FORCE Technology company for providing the X-ray lab equipment and their support to do the tests.

Funders	Funder number
Innovationsfonden

Keywords

Computer vision
Defect recognition
Industrial wool
X-ray

Publication forum classification

Publication forum level 1

ASJC Scopus subject areas

Computer Science Applications
Information Systems

Access to Document

10.1109/INDIN51400.2023.10217993

Cite this

Rafiei, M., Tran, D. T., & Iosifidis, A. (2023). Recognition of Defective Mineral Wool Using Pruned ResNet Models. In H. Dorksen, S. Scanzio, J. Jasperneite, L. Wisniewski, K. F. Man, T. Sauter, L. Seno, H. Trsek, & V. Vyatkin (Eds.), 2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023 (IEEE International Conference on Industrial Informatics (INDIN); Vol. 2023-July). IEEE. https://doi.org/10.1109/INDIN51400.2023.10217993

Rafiei, Mehdi ; Tran, Dat Thanh ; Iosifidis, Alexandros. / Recognition of Defective Mineral Wool Using Pruned ResNet Models. 2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023. editor / Helene Dorksen ; Stefano Scanzio ; Jurgen Jasperneite ; Lukasz Wisniewski ; Kim Fung Man ; Thilo Sauter ; Lucia Seno ; Henning Trsek ; Valeriy Vyatkin. IEEE, 2023. (IEEE International Conference on Industrial Informatics (INDIN)).

@inproceedings{59b7be94f77c413ea3218a3b8d5a678d,

title = "Recognition of Defective Mineral Wool Using Pruned ResNet Models",

abstract = "Mineral wool production is a non-linear process that makes it hard to control the final quality. Therefore, having a nondestructive method to analyze the product quality and recognize defective products is critical. For this purpose, we developed a visual quality control system for mineral wool. X-ray images of wool specimens were collected to create a training set of defective and non-defective samples. Afterward, we developed several recognition models based on the ResNet architecture to find the most efficient model. In order to have a light-weight and fast inference model for real-life applicability, two structural pruning methods are applied to the classifiers. Considering the low quantity of the dataset, cross-validation and augmentation methods are used during the training. As a result, we obtained a model with more than 98% accuracy, which in comparison to the current procedure used at the company, it can recognize 20% more defective products.",

keywords = "Computer vision, Defect recognition, Industrial wool, X-ray",

author = "Mehdi Rafiei and Tran, {Dat Thanh} and Alexandros Iosifidis",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; IEEE International Conference on Industrial Informatics ; Conference date: 17-07-2023 Through 20-07-2023",

year = "2023",

doi = "10.1109/INDIN51400.2023.10217993",

language = "English",

series = "IEEE International Conference on Industrial Informatics (INDIN)",

publisher = "IEEE",

editor = "Helene Dorksen and Stefano Scanzio and Jurgen Jasperneite and Lukasz Wisniewski and Man, {Kim Fung} and Thilo Sauter and Lucia Seno and Henning Trsek and Valeriy Vyatkin",

booktitle = "2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023",

address = "United States",

}

Rafiei, M, Tran, DT & Iosifidis, A 2023, Recognition of Defective Mineral Wool Using Pruned ResNet Models. in H Dorksen, S Scanzio, J Jasperneite, L Wisniewski, KF Man, T Sauter, L Seno, H Trsek & V Vyatkin (eds), 2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023. IEEE International Conference on Industrial Informatics (INDIN), vol. 2023-July, IEEE, IEEE International Conference on Industrial Informatics, Lemgo, Germany, 17/07/23. https://doi.org/10.1109/INDIN51400.2023.10217993

Recognition of Defective Mineral Wool Using Pruned ResNet Models. / Rafiei, Mehdi; Tran, Dat Thanh; Iosifidis, Alexandros.
2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023. ed. / Helene Dorksen; Stefano Scanzio; Jurgen Jasperneite; Lukasz Wisniewski; Kim Fung Man; Thilo Sauter; Lucia Seno; Henning Trsek; Valeriy Vyatkin. IEEE, 2023. (IEEE International Conference on Industrial Informatics (INDIN); Vol. 2023-July).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

TY - GEN

T1 - Recognition of Defective Mineral Wool Using Pruned ResNet Models

AU - Rafiei, Mehdi

AU - Tran, Dat Thanh

AU - Iosifidis, Alexandros

PY - 2023

Y1 - 2023

N2 - Mineral wool production is a non-linear process that makes it hard to control the final quality. Therefore, having a nondestructive method to analyze the product quality and recognize defective products is critical. For this purpose, we developed a visual quality control system for mineral wool. X-ray images of wool specimens were collected to create a training set of defective and non-defective samples. Afterward, we developed several recognition models based on the ResNet architecture to find the most efficient model. In order to have a light-weight and fast inference model for real-life applicability, two structural pruning methods are applied to the classifiers. Considering the low quantity of the dataset, cross-validation and augmentation methods are used during the training. As a result, we obtained a model with more than 98% accuracy, which in comparison to the current procedure used at the company, it can recognize 20% more defective products.

AB - Mineral wool production is a non-linear process that makes it hard to control the final quality. Therefore, having a nondestructive method to analyze the product quality and recognize defective products is critical. For this purpose, we developed a visual quality control system for mineral wool. X-ray images of wool specimens were collected to create a training set of defective and non-defective samples. Afterward, we developed several recognition models based on the ResNet architecture to find the most efficient model. In order to have a light-weight and fast inference model for real-life applicability, two structural pruning methods are applied to the classifiers. Considering the low quantity of the dataset, cross-validation and augmentation methods are used during the training. As a result, we obtained a model with more than 98% accuracy, which in comparison to the current procedure used at the company, it can recognize 20% more defective products.

KW - Computer vision

KW - Defect recognition

KW - Industrial wool

KW - X-ray

U2 - 10.1109/INDIN51400.2023.10217993

DO - 10.1109/INDIN51400.2023.10217993

M3 - Conference contribution

AN - SCOPUS:85171177167

T3 - IEEE International Conference on Industrial Informatics (INDIN)

BT - 2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023

A2 - Dorksen, Helene

A2 - Scanzio, Stefano

A2 - Jasperneite, Jurgen

A2 - Wisniewski, Lukasz

A2 - Man, Kim Fung

A2 - Sauter, Thilo

A2 - Seno, Lucia

A2 - Trsek, Henning

A2 - Vyatkin, Valeriy

PB - IEEE

T2 - IEEE International Conference on Industrial Informatics

Y2 - 17 July 2023 through 20 July 2023

ER -

Rafiei M, Tran DT, Iosifidis A. Recognition of Defective Mineral Wool Using Pruned ResNet Models. In Dorksen H, Scanzio S, Jasperneite J, Wisniewski L, Man KF, Sauter T, Seno L, Trsek H, Vyatkin V, editors, 2023 IEEE 21st International Conference on Industrial Informatics, INDIN 2023. IEEE. 2023. (IEEE International Conference on Industrial Informatics (INDIN)). doi: 10.1109/INDIN51400.2023.10217993

Recognition of Defective Mineral Wool Using Pruned ResNet Models

Abstract

Publication series

Conference

Funding

Keywords

Publication forum classification

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this