Lightweight AI for Cultural Pattern Recognition: Safeguarding South Asian Regional Embroidery Heritage

Abstract

The most important element of cultural heritage is the embroidery used in the ancient clothes. Embroidery used in the ancient clothes represents centuries of creative tradition, regional uniqueness, and the passing down of skills from one generation to the next. These elaborate textile practices, which have their roots in manual craftsmanship and oral instruction, are increasingly threatened by industrial production, globalization, and the slow decline of traditional artisans. By developing an automated identification structure based on MobileNetV2, an efficient yet highly effective convolutional neural network (CNN), this work presents an AI approach to support the preservation of regional embroidery. The model used in the paper was trained very carefully with a dataset of 5,200 high-resolution images that included a range of regional embroidery styles using pretrained ImageNet weights. With an identification accuracy of 97.3%, the suggested MobileNetV2 outperformed traditional CNN architectures like VGG16 and ResNet by over 2.3%. The results demonstrate how small AI models can make a significant contribution to the preservation of cultural heritage by providing a scalable method for storing and assessing textile arts for museums, scholars, and craftspeople. This methodology exhibits promise in fields such as digital archives, fashion technology, and cultural data analytics for heritage conservation.

Received: 17 August 2025 | Revised: 15 December 2025 | Accepted: 3 March 2026

Conflicts of Interest
The authors declare that they have no conflicts of interest to this work.

Data Availability Statement
The data that support the findings of this study are openly available in Kaggle at https://www.kaggle.com/datasets/ask1999/embroiderydataset.

Author Contribution Statement
Amir Sohail Khan: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Junjie Zhang: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration.