Textiles have an important role in many cultures and have been digitised. They are three-dimensional objects and have complex structures, especially archaeological fabric specimens and artifact textiles created manually by traditional craftsmen. In this article, we propose a novel algorithm for textile classification based on their structures. First, a hypergraph is used to represent the textile structure. Second, multisets of k-neighbourhoods are extracted from the hypergraph and converted to one feature vector for representation of each textile. Then, the k-neighbourhood vectors are classified using seven most popular supervised learning methods. Finally, we evaluate experimentally the different variants of our approach on a data set of 1,600 textile samples with the 4-fold cross-validation technique. The experimental results indicate that comparing the variants, the best classification accuracies are 0.999 with LR, 0.994 with LDA, 0.996 with KNN, 0.994 with CART, 0.998 with NB, 0.974 with SVM, and 0.999 with NNM.

An efficient classification algorithm for traditional textile patterns from different cultures based on structures

Nguyen Phuong Thanh
Writing – Original Draft Preparation
;
2021-01-01

Abstract

Textiles have an important role in many cultures and have been digitised. They are three-dimensional objects and have complex structures, especially archaeological fabric specimens and artifact textiles created manually by traditional craftsmen. In this article, we propose a novel algorithm for textile classification based on their structures. First, a hypergraph is used to represent the textile structure. Second, multisets of k-neighbourhoods are extracted from the hypergraph and converted to one feature vector for representation of each textile. Then, the k-neighbourhood vectors are classified using seven most popular supervised learning methods. Finally, we evaluate experimentally the different variants of our approach on a data set of 1,600 textile samples with the 4-fold cross-validation technique. The experimental results indicate that comparing the variants, the best classification accuracies are 0.999 with LR, 0.994 with LDA, 0.996 with KNN, 0.994 with CART, 0.998 with NB, 0.974 with SVM, and 0.999 with NNM.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11697/183215
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 1
social impact