Efficiency of Temporal Convolutional Networks in Karate Kata Evaluation: A Comparative Study

Viona Zatil Aqmar Kaleb, Wiranto Herry Utomo

Abstract


The objective quantification of complex martial-arts movement is a long-standing challenge in computer vision. This study reports a feasibility prototype that pairs MoveNet Lightning pose estimation with a lightweight Temporal Convolutional Network for two tasks on a small custom dataset of four foundational Karate Kata (Heian Shodan through Heian Yondan): (1) multi-class kata recognition and (2) binary correctness evaluation. A 132-dimensional kinematic feature vector is constructed per frame from 17 MoveNet keypoints, combining hip-centered normalized coordinates, confidence scores, joint angles, bone-length ratios, and end-effector velocities. The full dataset comprises 100 lateral-view videos (80 train/10 validation/10 test) collected from a single Indonesian dojo. A multi-task model with two causal-dilated 1D-convolution layers (32 filters, dilations 1 and 2) is trained for up to 100 epochs on Apple M1 hardware over five random seeds. The kata head reaches 100% validation accuracy with zero variance across all five seeds; however, this result is interpreted with strong reservations because the validation set contains only 10 samples. The correctness head consistently collapses to majority-class prediction (60.0% ± 0.0%, identical to the all-positive baseline of 6/10). The contribution of this paper is therefore a fully reproducible end-to-end pipeline (MoveNet → 132-D features → multi-task TCN → real-time webcam demo) together with a candid characterization of the dataset-driven limits that block the correctness task at this scale.

Keywords


Temporal Convolutional Network; Karate Kata; Deep Learning Efficiency; Pose Estimation; MoveNet;

Full Text:

PDF

References


Abdulrahman, N., Zainuddin, Z., & Nurtanio, I. (2026). A deep learning-based framework using CNN+LSTM for karate kata classification and correctness evaluation. Engineering, Technology & Applied Science Research, 16(1), 32619–32624. https://doi.org/10.48084/etasr.12147

Armstrong, K., Rodrigues, A., Willmott, A. P., Zhang, L., & Ye, X. (2025). Validation of human pose estimation and human mesh recovery for extracting clinically relevant motion data from videos. arXiv. https://doi.org/10.48550/arXiv.2503.14760

Bai, S., Kolter, J. Z., & Koltun, V. (2018). An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv. https://doi.org/10.48550/arXiv.1803.01271

Chen, J., Wang, J., Yuan, Q., & Yang, Z. (2023). CNN-LSTM model for recognizing video-recorded actions performed in a traditional Chinese exercise. IEEE Journal of Translational Engineering in Health and Medicine, 11, 351–359. https://doi.org/10.1109/JTEHM.2023.3282245

Cornman, H. L., Stenum, J., & Roemmich, R. T. (2021). Video-based quantification of human movement frequency using pose estimation: A pilot study. PLoS ONE, 16(12), e0261450. https://doi.org/10.1371/journal.pone.0261450

Ding, X., Li, Z., Yu, J., Xie, W., Li, X., & Jiang, T. (2023). A novel lightweight human activity recognition method via L-CTCN. Sensors, 23(24), 9681. https://doi.org/10.3390/s23249681

Echeverria, J., & Santos, O. C. (2021). Toward modeling psychomotor performance in karate combats using computer vision pose estimation. Sensors, 21(24), 8378. https://doi.org/10.3390/s21248378

Ercolano, G., & Rossi, S. (2021). Combining CNN and LSTM for activity of daily living recognition with a 3D matrix skeleton representation. Intelligent Service Robotics, 14(2), 175–185. https://doi.org/10.1007/s11370-021-00358-7

Fukushima, T., Blauberger, P., Guedes Russomanno, T., & Lames, M. (2025). Comparison of different pose estimation models for lower-body kinematics: A validation study. Scientific Journal of Sport and Performance, 5(2), 253–268. https://doi.org/10.55860/XWJL7156

Gupta, A., Gupta, K., Gupta, K., & Gupta, K. (2021). Human activity recognition using pose estimation and machine learning algorithm. Proceedings of the 2nd International Semantic Intelligence Conference (ISIC 2021), 2786, 323–330. https://ceur-ws.org/Vol-2786/Paper40.pdf

He, J.-L., Wang, J.-H., Lo, C.-M., & Jiang, Z. (2025). Human activity recognition via attention-augmented TCN-BiGRU fusion. Sensors, 25(18), 5765. https://doi.org/10.3390/s25185765

Islam, M. M., Nooruddin, S., Karray, F., & Muhammad, G. (2022). Human activity recognition using tools of convolutional neural networks: A state of the art review, data sets, challenges, and future prospects. Computers in Biology and Medicine, 149, 106060. https://doi.org/10.1016/j.compbiomed.2022.106060

Kim, J.-W., Choi, J.-Y., Ha, E.-J., & Choi, J.-H. (2023). Human pose estimation using MediaPipe Pose and optimization method based on a humanoid model. Applied Sciences, 13(4), 2700. https://doi.org/10.3390/app13042700

Ma, N., Zhang, X., Zheng, H.-T., & Sun, J. (2018). ShuffleNet V2: Practical guidelines for efficient CNN architecture design. In V. Ferrari, M. Hebert, C. Sminchisescu, & Y. Weiss (Eds.), Computer Vision – ECCV 2018 (pp. 122–138). Springer. https://doi.org/10.1007/978-3-030-01264-9_8

Malik, N. U. R., Abu-Bakar, S. A. R., Sheikh, U. U., Channa, A., & Popescu, N. (2023). Cascading pose features with CNN-LSTM for multiview human action recognition. Signals, 4(1), 40–55. https://doi.org/10.3390/signals4010002

Mortezapour Shiri, F., Yamaguchi, S., & Ahmadon, M. A. B. (2025). A deep learning model based on bidirectional temporal convolutional network (Bi-TCN) for predicting employee attrition. Applied Sciences, 15(6), 2984. https://doi.org/10.3390/app15062984

Priya, B. G., & Arulselvi, D. M. (2018). Action classification for karate dataset using deep learning. 2018 International Conference on Intelligent Computing and Communication for Smart World (I2C2SW), 1–4. https://doi.org/10.1109/I2C2SW45816.2018.8997335

Saeed, S. M., Akbar, H., Nawaz, T., Elahi, H., & Khan, U. S. (2023). Body-pose-guided action recognition with convolutional long short-term memory (LSTM) in aerial videos. Applied Sciences, 13(16), 9384. https://doi.org/10.3390/app13169384

Sekaran, S. R., Pang, Y. H., Ling, G. F., & Yin, O. S. (2022). MSTCN: A multiscale temporal convolutional network for user independent human activity recognition. F1000Research, 11, 1227. https://doi.org/10.12688/f1000research.126090.2

Shuvo, M. R., Mekala, M. S., & Elyan, E. (2025). Deep learning and attention-based methods for human activity recognition and anticipation: A comprehensive review. Cognitive Computation, 17(6), 158. https://doi.org/10.1007/s12559-025-10513-2

Washabaugh, E. P., Shanmugam, T. A., Ranganathan, R., & Krishnan, C. (2022). Comparing the accuracy of open-source pose estimation methods for measuring gait kinematics. Gait & Posture, 97, 188–195. https://doi.org/10.1016/j.gaitpost.2022.08.008

Yao, S., Ping, Y., Yue, X., & Chen, H. (2025). Graph convolutional networks for multi-modal robotic martial arts leg pose recognition. Frontiers in Neurorobotics, 18, 1520983. https://doi.org/10.3389/fnbot.2024.1520983

Ye, R. Z., Subramanian, A., Diedrich, D., Lindroth, H., Pickering, B., & Herasevich, V. (2022). Effects of image quality on the accuracy human pose estimation and detection of eye lid opening/closing using OpenPose and DLib. Journal of Imaging, 8(12), 330. https://doi.org/10.3390/jimaging8120330




DOI: http://dx.doi.org/10.30872/jim.v21i1.28888

Refbacks

  • There are currently no refbacks.


Copyright (c) 2026 Informatika Mulawarman : Jurnal Ilmiah Ilmu Komputer

Editor Informatika Mulawarman Address:
ISSN 1858-4853 (Print) | ISSN 2597-4963 (Online)
Published by: Mulawarman University
Managed by : Informatika Department
Jalan Sambaliung No.9 Sempaja Selatan Samarinda Utara,
Kalimantan Timur 75117 - Indonesia
E-mail: jim.unmul@gmail.com
OJS: http://e-journals.unmul.ac.id/index.php/JIM
Contact Person: Gubtha Mahendra Putra

 Creative Commons License

Informatika Mulawarman by http://e-journals.unmul.ac.id/index.php/JIM/index is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Under the CC BY-SA license, authors and other users are able to reprint, distribute or use the material for commercial purposes so long as they give attribution to the journal Informatika Mulawarman and license the republished material under the same license.