Yoga Pose Prediction Using InceptionV3 an Transfer Learning Approach

Baskar, Monika Maheswari (2024) Yoga Pose Prediction Using InceptionV3 an Transfer Learning Approach. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (733kB) | Preview
Preview	PDF (Configuration Manual) Download (640kB) | Preview

Abstract

This study recognizes human kinematics most accurately to help people have a good and healthy life in future by recognising the human posture using the InceptionV3 algorithm and Transfer learning. The field of computer vision has a difficult problem while estimating human posture error in posture can lead to discomfort and immediate issues. Identifying a person's posture in the photograph might be a little bit difficult due to the image size, resolution, lighting, background clutter, apparel and environment. The prediction is based on human skeleton motion through joint position using Mediapipe which detects the key points of body, foot, hand and face from single image. In order to predict whether the yoga postures are correct or wrong with the input image, the input image is changed to skeleton to get the performance metrics accurately. This research is fully based on predicting the yoga posture whether it is correct or incorrect by maximizing angles of 38 different yoga asanas through accurate prediction. This study in cooperates the CROSS-INDUSTRY STANDARD PROCESS FOR DATA MINING Methodology. The dataset was taken from Kaggle consists of 38 yoga asanas after doing the pre-processing After the pre- processing steps the images are augmented and converted into skeleton keyframes using the OpenCV and Mediapipe. The model used to predict the data is the InceptionV3 model which is a transfer learning approach and jupyter notebook is used for implementation. The models performance was evaluated using metrics such as accuracy, precision, recall and F1-score the training and validation accuracy was achieved as 96.03% and it also compared the STF-ResNet, AdaBoost, InceptionV3 and VGG model with accuracy around 77% to 93% respectively. Also, to provide the evaluation more compressively confusion matrix and classification report is provided. In recent years, yoga has been practiced by many people in terms of relaxation and for a peaceful mind but the problem with yoga is that like any other activity it is critical to do it because poor posture can make a yoga session unproductive and even detrimental. The proposed system aims to provide a comprehensive solution for individuals looking to enhance their yoga practice and improve their overall health and wellbeing.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Tomer, Vikas UNSPECIFIED
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 > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine > Personal Health and Hygiene
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	07 Aug 2025 10:28
Last Modified:	07 Aug 2025 10:28
URI:	https://norma.ncirl.ie/id/eprint/8462

Actions (login required)

View Item