Lightweight Method for Yoga Posture Recognition: Contributions to Well-Being and Quality of Life

• Caio C. M. Antunes UFAM
• Rafael C. Carvalho UFAM
• Bernardo B. Gatto UFAM
• Juan G. Colonna UFAM

DOI: https://doi.org/10.5753/sbcas.2025.6971

Resumo

Recognizing the growing importance of yoga for enhancing physical health and mental well-being, this paper proposes a lightweight neural network method for the automatic recognition of yoga postures from images. By leveraging skeletal keypoints, our model achieves efficient and accurate posture classification. We evaluated our approach on the Yoga-82 dataset using two data augmentation strategies: horizontal flipping of images and data balancing via random Gaussian noise addition combined with keypoint fusion. Our model attains an accuracy of 90.31% with only 85,582 parameters, demonstrating competitive performance relative to more resource-intensive methods. This efficiency makes the approach particularly suitable for resource-constrained environments, such as smartphones, and paves the way for developing tutor applications that promote individual yoga practice and enhance overall well-being.

Referências

Abowd, D., Dey, A. K., Orr, R., and Brotherton, J. (1998). Contextawareness in wearable and ubiquitous computing. Virtual Reality, 3:200–211.

Anilkumar, A., KT, A., Sajan, S., and Sreekumar, S. (2021). Pose estimated yoga monitoring system. SSRN Electronic Journal.

Ashraf, F. B., Islam, M. U., Kabir, M. R., and Uddin, J. (2023). Yonet: A neural network for yoga pose classification. SN Computer Science, 4(2):198.

Chiddarwar, G. G., Ranjane, A., Sinha, M., and Gupta, A. (2020). Ai-based yoga pose estimation for android applications. International Journal of Engineering Research, 9:34–42.

Chollet, F. et al. (2015). Keras. [link].

Ek, S., Portet, F., and Lalanda, P. (2023). Transformer-based models to deal with heterogeneous environments in human activity recognition. Personal and Ubiquitous Computing, 27(6):2267–2280.

Garg, S., Saxena, A., and Gupta, R. (2022). Yoga pose classification: A cnn and mediapipe inspired deep learning approach for real-world application. Journal of Ambient Intelligence and Humanized Computing, pages 1–12.

Kothari, S. (2020). Yoga Pose Classification Using Deep Learning. Master’s thesis, San Jose State University.

Kumar, V., Choudhary, A., and Cho, E. (2020). Data augmentation using pre-trained transformer models. arXiv preprint arXiv:2003.02245.

Kunze, K., Barry, M., Heinz, E. A., Lukowicz, P., Majoe, D., and Gutknecht, J. (2006). Towards recognizing tai chi-an initial experiment using wearable sensors. In 3rd International Forum on Applied Wearable Computing 2006, pages 1–6. VDE.

Liu, J., Wang, G., Duan, L.-Y., Abdiyeva, K., and Kot, A. C. (2017). Skeleton-based human action recognition with global context-aware attention lstm networks. IEEE Transactions on Image Processing, 27(4):1586–1599.

Lugaresi, C. et al. (2019). Mediapipe: A framework for building perception pipelines. arXiv preprint arXiv:1906.08172.

Maurtua, I., Kirisci, P. T., Stiefmeier, T., Sbodio, M. L., and Witt, H. (2007). A wearable computing prototype for supporting training activities in automotive production. In 4th International Forum on Applied Wearable Computing 2007, pages 1–12. VDE.

Mazzia, V., Angarano, S., Salvetti, F., Angelini, F., and Chiaberge, M. (2022). Action transformer: A self-attention model for short-time pose-based human action recognition. Pattern Recognition, 124:108487.

Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., and Chen, L.-C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4510–4520.

Shavit, Y. and Klein, I. (2021). Boosting inertial-based human activity recognition with transformers. IEEE Access, 9:53540–53547.

Swain, D., Satapathy, S., Mishra, B., and Pati, S. K. (2022). Deep learning models for yoga pose monitoring. Algorithms, 15(11):403.

Uddin, S., Nawaz, T., Ferryman, J., Rashid, N., Asaduzzaman, M., and Nawaz, R. (2024). Skeletal keypoint-based transformer model for human action recognition in aerial videos. IEEE Access.

Vaswani, A. (2017). Attention is all you need. Advances in Neural Information Processing Systems.

Verma, M., Kumawat, S., Nakashima, Y., and Raman, S. (2020). Yoga-82: A new dataset for fine-grained classification of human poses. In IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pages 4472–4479.

Vrigkas, M., Nikou, C., and Kakadiaris, I. A. (2015). A review of human activity recognition methods. Frontiers in Robotics and AI, 2:28.

Woznowski, P., King, R., Harwin, W., and Craddock, I. (2016). A human activity recognition framework for healthcare applications: ontology, labelling strategies, and best practice. In International Conference on Internet of Things and Big Data, volume 2, pages 369–377. SciTePress.

Yadav, S. K., Shukla, A., Tiwari, K., Pandey, H. M., and Akbar, S. A. (2023). An efficient deep convolutional neural network model for yoga pose recognition using single images. arXiv preprint arXiv:2306.15768.