Color Feature Extraction and Clustering Map Construction of Bai Women’s Traditional Costumes in Dali Using Deep Neural Networks

The colors of ethnic costumes encode rich cultural meanings, yet conventional approaches often struggle to extract color characteristics objectively and quantify relationships among colors. This study develops a digital analysis framework integrating deep neural networks, K-means clustering, and color network modeling to examine Bai women’s traditional costumes in Dali. A total of 210 valid samples (n = 210) were collected and selected, covering three regional types (Erhai lakeside, basin, and mountainous) and two groups (young and middle-aged, and elderly). Color and visual features were extracted from costume images using deep convolutional neural networks, and geographic-region and age-group classification tasks were designed to test their ability to distinguish geographical origins and group membership. Unsupervised clustering was then applied to the high-dimensional color features, while a color network model was used to characterize the structure of color combinations through measures including color co-occurrence and node centrality. Models incorporating the color modality achieved accuracies of 83.33–88.89% in regional classification and 68.42–78.95% in age-group classification, indicating that color features can help distinguish the groups. Clustering revealed clear regional differences, suggesting possible associations between color choices, natural environments, and livelihoods, as well as distinct color combinations between younger and older groups. Network analysis identified stable, multilayered color combinations and group differences in core-color substitutability and the diversity of auxiliary-color clusters. The integrated sequence of feature extraction, classification validation, clustering, and network modeling offers a reusable framework for quantitatively comparing regional and group differences and supports the digital preservation and systematic documentation of traditional costume colors.

Authors

Institutions

Publication Details

Journal
Heritage
Published
2026-11-05
DOI
https://doi.org/10.3390/heritage9110463
Primary Topic
Fashion and Cultural Textiles
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Color Feature Extraction and Clustering Map Construction of Bai Women’s Traditional Costumes in Dali Using Deep Neural Networks

Yan Gao, Qingxia Fang, Shanxiong Chen, Jiahao Zhang
Heritage
Fashion and Cultural Textiles
article

Color Feature Extraction and Clustering Map Construction of Bai Women’s Traditional Costumes in Dali Using Deep Neural Networks

Yan Gao, Qingxia Fang, Shanxiong Chen, Jiahao Zhang
article en

Abstract

The colors of ethnic costumes encode rich cultural meanings, yet conventional approaches often struggle to extract color characteristics objectively and quantify relationships among colors. This study develops a digital analysis framework integrating deep neural networks, K-means clustering, and color network modeling to examine Bai women’s traditional costumes in Dali. A total of 210 valid samples (n = 210) were collected and selected, covering three regional types (Erhai lakeside, basin, and mountainous) and two groups (young and middle-aged, and elderly). Color and visual features were extracted from costume images using deep convolutional neural networks, and geographic-region and age-group classification tasks were designed to test their ability to distinguish geographical origins and group membership. Unsupervised clustering was then applied to the high-dimensional color features, while a color network model was used to characterize the structure of color combinations through measures including color co-occurrence and node centrality. Models incorporating the color modality achieved accuracies of 83.33–88.89% in regional classification and 68.42–78.95% in age-group classification, indicating that color features can help distinguish the groups. Clustering revealed clear regional differences, suggesting possible associations between color choices, natural environments, and livelihoods, as well as distinct color combinations between younger and older groups. Network analysis identified stable, multilayered color combinations and group differences in core-color substitutability and the diversity of auxiliary-color clusters. The integrated sequence of feature extraction, classification validation, clustering, and network modeling offers a reusable framework for quantitatively comparing regional and group differences and supports the digital preservation and systematic documentation of traditional costume colors.

HeritageVol. 9(11)
Southwest University (CN)
Openalex Percentile: Top 3%
Fashion and Cultural Textiles
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon