NORMA eResearch @NCI Library

Exercise detection and tracking using MediaPipe BlazePose and Spatial-Temporal Graph Convolutional Neural Network

Raju, Krishnanunni (2022) Exercise detection and tracking using MediaPipe BlazePose and Spatial-Temporal Graph Convolutional Neural Network. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (2MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


The purpose of this study is to create a deep-learning model based on a pose estimation framework and the Spatial-temporal Graph Convolutional Network that can track and classify exercises performed by humans into different categories. An open-source dataset available online is used to train the model. Despite being a well-liked posture estimation framework, OpenPose struggles while the subject is moving quickly and cannot cope up with cameras with fps greater than 22. This restriction is overcome by using MediaPipe BlazePose. It is used to extract the skeleton information from a moving individual and provides 33 key points of the body. Once a human is identified, the subject is tracked, feature extraction is carried out, and classification is done. The time taken to perform each exercise is recorded. The ST-GCN model implemented in this research was evaluated using accuracy of top-1% and accuracy of top-5%. Four variations of ST-GCN was implemented and evaluated in this research. The best model was obtained when a partitioning strategy of spatial configuration along with learnable edge importance weighting was used for ST-GCN. This model achieved a top-1 accuracy of 41.75% and top-5 accuracy of 89.32%. This solution will to an extend eliminate the need for another person or sensors attached to the body to keep track of the workout.

Item Type: Thesis (Masters)
Subjects: G Geography. Anthropology. Recreation > GV Recreation Leisure
Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QP Physiology
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 01 Mar 2023 15:14
Last Modified: 01 Mar 2023 17:27

Actions (login required)

View Item View Item