Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/227570
Title: FABRIC DEFECT DETECTION WITH DATA-DRIVEN MODELS
Authors: ZHANG CHEN
ORCID iD:   orcid.org/0000-0001-8469-6153
Keywords: Fabric defects, Defect detection, Data-driven model, Object Detector, Deep learning
Issue Date: 30-Mar-2022
Citation: ZHANG CHEN (2022-03-30). FABRIC DEFECT DETECTION WITH DATA-DRIVEN MODELS. ScholarBank@NUS Repository.
Abstract: Fabric defects bother the clothes manufacturers by causing a considerable loss. A garment factory sent us some textile samples and expected us to offer an automated defect detection strategy to maximize their profit. Hereby, deep learning is proposed to solve the real engineering problem. Data-driven models can identify the fabric defect in real-time with reasonable accuracy. In this project, the best Average Precision (AP) is 31.01% in detecting fabric defects. Two object detection models are implemented in the project. The first model is a variant of Faster Region Proposal-Convolutional Neural Network (Faster R-CNN), modified with a Feature Pyramid Network (FPN). The second model is Retinal Convolutional Neural Network (RetinaNet), a one-stage detector that has a simpler architecture. A well-labeled dataset with five major classes of defects is created carefully in Common Objects in Context (COCO) format. After training, RetinaNet acquires a shorter inference time but also good generalization performance.
URI: https://scholarbank.nus.edu.sg/handle/10635/227570
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
ZHANG CHEN(ZhangC).pdf2.02 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

30
checked on Sep 29, 2022

Download(s)

4
checked on Sep 29, 2022

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.