Klasifikasi Gerakan Tari Bali Perempuan Menggunakan Metode Spatial-Temporal Graph Convolutional Network (ST-GCN)

Daniel Sande Bona

doi:10.58602/jaiti.v4i2.253

Daniel Sande Bona ^{(Corresponding Author)} Institut Seni Budaya Indonesia Tanah Papua

DOI: https://doi.org/10.58602/jaiti.v4i2.253

Keywords: ST-GCN, Tari Bali, Skeleton, pengenalan aksi, deep learning, MediaPipe Pose

Abstract

Pengenalan gerakan tari Bali berpotensi mendukung dokumentasi warisan budaya, media pembelajaran, dan sistem umpan balik gerak berbasis komputer. Namun, model klasifikasi video RGB dapat mempelajari latar belakang, kostum, pencahayaan, atau identitas penari sebagai shortcut, bukan struktur gerakan. Penelitian ini bertujuan menyusun baseline Spatial-Temporal Graph Convolutional Network (ST-GCN) berbasis skeleton untuk klasifikasi 13 gerakan dasar tari Bali perempuan menggunakan MediaPipe Pose. Setiap video diproses menjadi 33 landmark tubuh dengan kanal koordinat x, koordinat y, dan visibility, kemudian distandarkan menjadi 64 frame. Folder train dan validation asli digabungkan hanya untuk validasi silang 5-fold berstratifikasi, sedangkan folder test resmi dipertahankan sebagai holdout akhir. Model menggunakan graf 33 landmark MediaPipe, backbone ST-GCN, dan GCNHead dengan global pooling serta linear classifier. Hasil validasi silang memperoleh top-1 accuracy 99,55% +/- 0,49%, top-5 accuracy 99,92% +/- 0,17%, dan macro F1 99,49% +/- 0,55%. Evaluasi holdout akhir satu kali menghasilkan top-1 accuracy 99,39%, top-5 accuracy 100,00%, dan macro F1 99,39%. Audit duplikasi identifier dan overlap SHA-256 tidak menemukan kebocoran data train-test. Strategi evaluasi ini menegaskan pemisahan antara validasi model dan pengujian akhir. Hasil ini menunjukkan bahwa ST-GCN berbasis skeleton menjadi baseline within-dataset yang kuat untuk pengenalan gerak tari Bali, meskipun generalisasi subject-independent belum dapat diklaim karena dataset tidak menyediakan identitas penari.

Downloads

Download data is not yet available.

References

I. P. P. B. Lantara, I. P. D. Payana, G. A. S. Pratama, W. E. A. Munayana, K. S. Nopiani, and I. N. R. Hendrawan, "GETARI: Dataset untuk klasifikasi gerakan dasar Tari Bali perempuan," Jurnal Nasional Pendidikan Teknik Informatika: JANAPATI, vol. 11, no. 3, pp. 290-300, 2022, doi: 10.23887/janapati.v11i3.52598.

I. N. R. Hendrawan, P. Setyarini, A. T. Tedja, I. P. P. B. Lantara, I. P. D. Payana, G. A. S. Pratama, W. E. A. Munayana, and K. S. Nopiani, "Video Dataset of Woman Basic Balinese Dance Movement for Action Recognition," Mendeley Data, version 2, 2023, doi: 10.17632/s2gv9d6gpb.2.

N. W. Utami, M. A. P. Putra, I. G. J. E. Putra, and G. A. Sampedro, "A CNN-based information system for Balinese dance classification with hyperparameter optimization," International Journal of Advances in Data and Information Systems, vol. 6, no. 2, pp. 376-390, 2025, doi: 10.59395/ijadis.v6i2.1402.

C. Irawan, H. P. Hadi, C. Jatmoko, and M. Doheir, "Video classification of Indonesian traditional dance using a hybrid CNN-LSTM model with pose estimation," Bulletin of Electrical Engineering and Informatics, vol. 15, no. 1, pp. 787-798, 2026, doi: 10.11591/eei.v15i1.11093.

S. Yan, Y. Xiong, and D. Lin, "Spatial temporal graph convolutional networks for skeleton-based action recognition," in Proc. AAAI Conference on Artificial Intelligence, vol. 32, no. 1, 2018, doi: 10.1609/aaai.v32i1.12328.

H. Duan, Y. Zhao, K. Chen, D. Lin, and B. Dai, "Revisiting skeleton-based action recognition," in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2022, pp. 2969-2978.

K. Guo, P. Wang, P. Shi, C. He, and C. Wei, "A new partitioned spatial-temporal graph attention convolution network for human motion recognition," Applied Sciences, vol. 13, no. 3, Art. no. 1647, 2023, doi: 10.3390/app13031647.

M. H. Al-Hakimi, I. Ahmed, M. Haseeb, T. H. Rassem, F. H. Quradaa, and R. S. Almoqbily, "An enhanced spatial-temporal graph convolution network with high order features for skeleton-based action recognition," PLOS ONE, vol. 20, no. 10, Art. no. e0332815, 2025, doi: 10.1371/journal.pone.0332815.

H. Hu, Y. Cao, Y. Fang, and Z. Meng, "Semantics-assisted training graph convolution network for skeleton-based action recognition," Sensors, vol. 25, no. 6, Art. no. 1841, 2025, doi: 10.3390/s25061841.

P. Jin, S. Guo, and C. Li, "Center-of-gravity-aware graph convolution for unsafe behavior recognition of construction workers," Sensors, vol. 25, no. 17, Art. no. 5493, 2025, doi: 10.3390/s25175493.

C. Lu, "Feature data analysis of dance movements by motion capture," Journal of Measurements in Engineering, vol. 13, no. 3, pp. 701-713, 2025, doi: 10.21595/jme.2025.24742.

X. Yan, J. Yang, and T. Salami, "Classification of Indian classical dances using MnasNet architecture with advanced polar fox optimization for hyperparameter optimization," Scientific Reports, vol. 15, Art. no. 3054, 2025.

M. Clayton, J. Li, A. Clarke, and M. Weinzierl, "Hindustani raga and singer classification using 2D and 3D pose estimation from video recordings," Journal of New Music Research, vol. 52, no. 4, pp. 285-300, 2024, doi: 10.1080/09298215.2024.2331788.

P. Malavath and N. Devarakonda, "Natya Shastra: Deep learning for automatic classification of hand mudra in Indian classical dance videos," Revue d Intelligence Artificielle, vol. 37, no. 3, pp. 251-260, 2023, doi: 10.18280/ria.370317.

T. Yoddamnern and P. Riyamongkol, "Deep learning for two-person Mae Mai Muay Thai classification: Leveraging normalized human pose sequences and CNN-LSTM," Scientific Culture, vol. 11, no. 4, pp. 333-349, 2025, doi: 10.5281/zenodo.11042528.

M. Muslih and E. H. Rachmawanto, "Convolutional Neural Network (CNN) untuk klasifikasi citra penyakit diabetes retinopathy," SKANIKA: Sistem Komputer dan Teknik Informatika, vol. 5, no. 2, pp. 167-176, 2022, doi: 10.36080/skanika.v5i2.2945.

A. S. Kusuma, A. I. Pradana, and B. W. Pamekas, "Pengembangan sistem perhitungan jumlah kendaraan berdasarkan jenis kendaraan menggunakan algoritma YOLO secara realtime," SKANIKA: Sistem Komputer dan Teknik Informatika, vol. 7, no. 2, pp. 166-179, 2024.

C. Lugaresi et al., "MediaPipe: A framework for building perception pipelines," arXiv:1906.08172, 2019.

MMAction2 Contributors, "OpenMMLab action understanding toolbox and benchmark," 2020. [Online]. Available: https://github.com/open-mmlab/mmaction2

R. Hendrawan, D. Payana, D. Payana, and A. Tedja, "rudyhendrawn/traditional-dance- videoclassification: v1.0.0," Zenodo, Mar. 13, 2023, doi: 10.5281/zenodo.7726950.