Efficiency marker evaluation based on optimized deep learning supported by Bayesian optimization technique

In this article the efficiency of marker patterns is calculated based on a new optimized automated computer-aided design system interested in nesting the irregular objects using deep learning techniques to reduce the unused area of garment layout. The automatic hyper-parameter is adjusted based on B...

Full description

Saved in:
Bibliographic Details
Published in:Textile research journal 2023-09, Vol.93 (17-18), p.4273-4289
Main Authors: Salama, Wessam M., Aly, Moustafa H., Abed, Ahmed M., Eladly, Amira M.
Format: Article
Language:English
Subjects:
Algorithms
Bayesian analysis
CAD
Computer aided design
Computing time
Data augmentation
Datasets
Decision trees
Deep learning
Design optimization
Efficiency
Machine learning
Nesting
Optimization techniques
Support vector machines
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this article the efficiency of marker patterns is calculated based on a new optimized automated computer-aided design system interested in nesting the irregular objects using deep learning techniques to reduce the unused area of garment layout. The automatic hyper-parameter is adjusted based on Bayesian optimization technique. Our proposed work is based on four phases: segmentation using the Grab-Cut technique which is utilized to increase our framework performance and decrease the cost. Moreover, different optimization algorithms, including k-nearest neighbor, naive Bayes, decision trees, and support vector machine are performed for classifying the pattern size from the fabric layer. Furthermore, the data augmentation technique is applied to overcome the lack of datasets and improve our framework performance by increasing the number of datasets. The optimized framework achieves 98.99% area under the receiver-operating characteristic curve, an accuracy of 98.97%, a sensitivity ( S e ) of 98.98%, a precision ( P r ) of 98.89%, an F 1 -score of 98.99%, a mean square error of 0.03% to 4.91%, an efficiency of 99.66%, a p-value of 0.0207, a t-statistics of 1.336, and a computational time of 3.21 s.
ISSN:0040-5175
1746-7748
DOI:10.1177/00405175231171720