NORMA eResearch @NCI Library

Identification and Classification of Defects in Steel Sheets using Deep Learning Models

Bansal, Akansha (2020) Identification and Classification of Defects in Steel Sheets using Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (1MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


Defect detection is one of the critical parts of manufacturing of a product in an industry. Manually inspecting the bulk production of steel sheets for defect detection is a tiresome task and could be prone to human errors. When steel sheets are manufactured it came in contact with many heavy machines for the purpose of drying, cutting, rolling, heating due to this some defects occur on the surface. So, this technical report aims at developing a deep learning model which can identify the surface defects in steel sheets and classify those defects into various classes. Automating this task can improve the manufacturing standards of organization. For this purpose, four deep learning models has been implemented on Severstal dataset. The models are Xception, U-Net, Mask RCNN and UNet++ and their performance has been compared by Dice Coefficient. Among all the four models Xception achieved highest Dice Coefficient of 0.927. In addition to this, results of the previous used techniques for defect detected were discussed. Based on the identified gaps deep learning models were developed.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Dan English
Date Deposited: 22 Jan 2021 10:30
Last Modified: 22 Jan 2021 10:30

Actions (login required)

View Item View Item