UTAR Institutional Repository

Automated visual defect detection using deep learning

Loh, Xiao (2022) Automated visual defect detection using deep learning. Final Year Project, UTAR.

[img]
Preview
PDF
Download (4Mb) | Preview

    Abstract

    A manufacturing defect is a flaw that causes a product to deviate from its intended design, thereby losing its quality and no longer having its due value. There are two kinds of methods employed by the manufacturing industries to ensure that the manufactured products are well-conditioned and free of any defects, namely human quality inspection and artificial intelligence visual inspection. However, the former faces many limitations and problems, resulting in low quality control efficiency, while the latter is relatively reliable and effective. Artificial intelligence visual inspection is a technique which utilises computer vision and deep learning technology to mechanically “see” a product and determine whether it has defects, without any human involvement. The main goal of this project is to study and develop various automated defect detection models by utilizing state-of-the-art deep learning segmentation algorithms, including U-Net, Double U-Net, SETR, TransU-Net, TransDAU-Net, CAM and SEAM to perform semantic segmentation in fully supervised and weakly supervised learning manners. Model analysis and evaluation are performed to compare the performance of all the algorithms in a variety of aspects. In this project, a magnetic tile defect dataset and a production item surface defect dataset are employed to train and evaluate deep learning segmentation models. By applying the models, the detected defects will be segmented and classified, and the output results will be displayed to the user through a coloured segmentation mask for each defect. Manufactures will be able to automate the industrial inspection process by implementing the deep learning models proposed in this project. Quality control procedures of the industries will be sublimated to another level by the automation of product defect detection.

    Item Type: Final Year Project / Dissertation / Thesis (Final Year Project)
    Subjects: T Technology > T Technology (General)
    Divisions: Faculty of Information and Communication Technology > Bachelor of Computer Science (Honours)
    Depositing User: ML Main Library
    Date Deposited: 11 Jan 2023 18:30
    Last Modified: 11 Jan 2023 18:30
    URI: http://eprints.utar.edu.my/id/eprint/4707

    Actions (login required)

    View Item