Related Experiment Video
Updated: Sep 27, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
4.2K
Research on Multiplayer Posture Estimation Technology of Sports Competition Video Based on Graph Neural Network
1Shaanxi Normal University, Xi'an, Shaanxi 710000, China.
Computational Intelligence and Neuroscience
|April 11, 2022
Summary
This study introduces a novel graph neural network approach for accurate human pose estimation in sports videos. The method effectively addresses limitations of traditional techniques, improving multi-person pose analysis in dynamic game footage.
Area of Science:
- Computer Vision
- Sports Analytics
- Machine Learning
Background:
- Traditional sports video analysis relies on manual annotation, which is costly and limited.
- Existing methods using manual features for human pose detection lack accuracy, especially with occlusions.
Purpose of the Study:
- To develop an advanced human pose detection method for sports videos.
- To improve the accuracy and efficiency of multi-person pose estimation in sports competitions.
Main Methods:
- Adapted a neural network, integrating local and global features inspired by Deep-ID.
- Combined the improved neural network with a human joint model.
- Developed a graph neural network-based approach for human pose detection.
Main Results:
- The proposed algorithm demonstrates superior human posture estimation in sports videos.
- Achieved strong performance in multi-person pose estimation tasks within sports game footage.
Conclusions:
- The graph neural network-based method offers a significant improvement over traditional approaches.
- This technique is effective for analyzing complex human poses in dynamic sports environments.
