Determining success in snatch weightlifting using pose landmarks and barbell detection

Name: Ming Qi

LCSH: Weight training

LCSH: Machine learning

LCSH: Weight lifting -- Coaching

Abstract: This research integrated computer vision and machine learning techniques to objectively evaluate snatch weightlifting success. By leveraging MediaPipe for skeletal detection and You Only Look Once (YOLO) object detection for barbell detection, the study classified snatch into six phases. An artificial neural network (ANN) and support vector machine (SVM) were applied to classify weightlifting phases from features extracted using MediaPipe. The distances between an athlete’s hands and a barbell were computed using the MediaPipe features, which represented the points on an athlete’s right and left hands as well as the points on a barbell. This study employed different methods to evaluate weightlifting success. For example, the method that used the holding period of the sixth phase could obtain a 95% accuracy rate, whereas the method that evaluated the presence of all six phases in sequence could derive a lower accuracy of 70%. A method that evaluated the ordered six phases, the holding time of the sixth phase, and the barbell slipping achieved the highest accuracy rate of 100%. The proposed method, which did not require specialized equipment, could achieve notable weightlift phase classification and efficiently determine the success or failure of snatch weightlifting

Rangsit University Library

Address: Pathumthani

Email: library@rsu.ac.th

Name: Rong Phoophuangpairoj

Role: Advisor

Created: 2024

Modified: 2569-01-18

Issued: 2026-01-18

วิทยานิพนธ์/Thesis

application/pdf

eng

DegreeName: Master of Engineering

Level: Master's Degree

Descipline: Electrical and Computer Engineering

Grantor: Rangsit University

©copyrights Rangsit University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	MING QI.pdf	1.17 MB

ใช้เวลา

0.163057 วินาที