Original scientific paper

https://doi.org/10.17559/TV-20241218002203

Development of a Performance Evaluation System in Turkish Folk Dance Using Deep Learning-Based Pose Estimation

Erdem Büyükgökoğlan ; Isparta University of Applied Sciences, Department of Computer Engineering, Isparta, Turkey
Sinan Uğuz orcid.org/0000-0003-4397-6196 ; Isparta University of Applied Sciences, Department of Computer Engineering, Isparta, Turkey *

* Corresponding author.

Abstract

Folk dances are integral cultural expressions that reflect a society's heritage, values, and historical development. In the context of education, folk dances promote belonging, discipline, and creativity. This study explores an innovative approach to teaching and evaluating traditional Turkish folk dance, through the integration of deep learning, time-series analysis, and classical methods. The system developed in this study enables objective performance assessment by students and provides teachers with valuable feedback for monitoring progress. The performances of one teacher and nine students, each performing five distinct dance figures, were captured via webcam recordings, and skeletal data were analysed using Mediapipe and YOLO. To evaluate performance, classical methods such as Euclidean distance and Cosine similarity were employed in conjunction with time-series techniques like TLCC and DTW, as well as deep learning models such as LSTM and Siamese Networks. Among these methods, the LSTM model emerged as the most effective, achieving an average score of 68,43 and an MSE of 56,11. The DTW method followed, achieving an average score of 60,64 and an MSE of 139,32. Overall, the integration of technology, particularly deep learning, into traditional dance education presents a transformative opportunity for enhancing performance evaluation in folk dance.

Keywords

dance evaluation, DTW, folk dance, LSTM, pose estimation, siamese networks, TLCC

Hrčak ID:

335067

URI

https://hrcak.srce.hr/335067

Publication date:

30.8.2025.

Visits: 390 *