RECOGNITION OF WORKOUT EXERCISE BASED ON IMAGE PROCESSING USING CNN MOBILENETV2 AND EFFICIENTNETB3

Main Article Content

Andrew Hendisutio
Meirista Wulandari
Wahidin Wahab

Abstract

Exercise is crucial for maintaining a healthy and fit body, yet many individuals struggle to track their workout progress effectively. This paper explores the application of deep learning techniques, specifically Convolutional Neural Networks (CNNs), to recognize and classify common exercise moves such as push-ups, pull-ups, sit-ups, leg raises, and squats. The goal is to provide individuals with a tool to assist in counting repetitions and sets, thereby enhancing their exercise experience. The study employs two well-known CNN models, MobileNetV2 and EfficientNetB3, trained on a custom dataset consisting of 2,500 exercise move images. The dataset includes various training-to-testing data ratios, ranging from 90:10 to 50:50. The models are evaluated based on their accuracy in classifying exercise moves, and confusion matrices are generated to analyze their performance further.

Article Details

Section
Articles

References

[1] UNICEF, “Landscape Analysis of Overweight and Obesity in Indonesia,” 2021. Accessed:Sep. 25, 2023. [Online]. Available:

https://www.unicef.org/indonesia/media/15481/file/Landscape%20analysis%20of%20overweight%20and%20obesity%20in%20Indonesia.pdf

[2] V. M. Iversen, M. Norum, B. J. Schoenfeld, and M. S. Fimland, “No Time to Lift? Designing Time-Efficient Training Programs for Strength and Hypertrophy: A Narrative Review,” Sports Medicine, vol. 51, no. 10. Springer Science and Business Media Deutschland GmbH, pp.

2079–2095, Oct. 01, 2021. doi: 10.1007/s40279-021-01490-1.

[3] K. Skawinski, F. M. Roca, and R. Dieter, “Workout Type Recognition and Repetition Counting with CNNs from 3D Acceleration Sensed on the Chest.”

[4] R. Arlin and R. Munir, “The Development of Push Up Counter Android Application with Computer Vision.”

[5] K. Soomro and A. R. Zamir, “Action recognition in realistic sports videos,” Advances in Computer Vision and Pattern Recognition, vol. 71, pp. 181–208, 2014, doi: 10.1007/978-3-319-09396-3_9.

[6] G. Li and C. Zhang, “Automatic detection technology of sports athletes based on image recognition technology,” EURASIP J Image Video Process, vol. 2019, no. 1, Dec. 2019, doi:10.1186/s13640-019-0415-x.

[7] Evan, M. Wulandari, and E. Syamsudin, “Recognition of pedestrian traffic light using tensorflow and SSD MobileNet V2,” in IOP Conference Series: Materials Science and Engineering, IOP Publishing Ltd, Dec. 2020. doi: 10.1088/1757-899X/1007/1/012022.

[8] M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” May 2019, [Online]. Available: http://arxiv.org/abs/1905.11946

[9] B. Schoenfeld et al., “Resistance Training Recommendations to Maximize Muscle Hypertrophy in an Athletic Population: Position Stand of the IUSCA,” International Journal of Strength and Conditioning, vol. 1, no. 1, Aug. 2021, doi: 10.47206/ijsc.v1i1.81.

[10] Hasyim Abdillah, “Workout/Exercise Images.” Accessed: Sep. 25, 2023. [Online]. Available: https://www.kaggle.com/datasets/hasyimabdillah/workoutexercises-images

[11] V. Tyagi, Understanding Digital Image Processing. CRC Press, 2018. doi: 10.1201/9781315123905.

[12] R. Indraswari, R. Rokhana, and W. Herulambang, “Melanoma image classification based on MobileNetV2 network,” in Procedia Computer Science, Elsevier B.V., 2021, pp. 198–207. doi: 10.1016/j.procs.2021.12.132.