Analisis Dataset Video Gerakan Dasar Taekwondo dengan Motion Capture

Sarinastiti, Widi; Setyowati, Aulia Dwi; Basuki, Achmad

doi:10.33005/sibc.v15i2.23

Cited by 1 publication

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After the program was modified by taking the z value and changing the angle calculation, taekwondo movement image data began to be collected as data used as movement references and data to test machine learning. The taekwondo movement data collected consists of nine basic taekwondo movements, namely joochoom seogi, apkoobi, kyorugi joonbi, arae jireugi, momtong jireugi, eolgoul jireugi, arae makki, arae hecho makki, and eolgoul makki (Sarinastiti, Setyowati, & Basuki, 2022). Then proceed with collecting the nine basic movements from a different point of view.…”

Section: Fig 1 Research Stages Flowchartmentioning

confidence: 99%

Development of Independent Taekwondo Training Machine Learning With 3d Pose Model Mediapipe

Santoso,

Sandjaya

2023

SinkrOn

View full text Add to dashboard Cite

Taekwondo is a martial art that focuses on punching and kicking movements while upholding the values of discipline, ethics, and good behavior. Discipline is built with routine training to make someone proficient in taekwondo martial arts. Training cannot be carried out flexibly because it must be accompanied by a sabouem to know the correct taekwondo movements. Machine learning can be used as a solution for taekwondo movement recognition by building a learning machine model that recognizes the correct taekwondo movement. The MediaPipe framework has the advantage of being able to recognize human posture with 33 points or landmarks. The research was carried out by conducting a literature study, where similar research was found but only based on the values of the x and y axes. So a problem arises where the majority of taekwondo movements require the z axis to know the correct taekwondo movements. The research was conducted to add z-axis values and change calculations, which were adjusted to reconstruct a training data object in the form of an image into a three-dimensional shape. From this study, it was found that machine learning using the x, y, and z axes is much better for its use, especially when detecting taekwondo movements from different viewpoints from the training data. This research can be developed by enlarging the image dataset and packaging the model into a mobile application so that it can be used for taekwondo training up to taekwondo movement assessment.

show abstract

Section: Fig 1 Research Stages Flowchartmentioning

confidence: 99%