Related Experiment Video
Updated: Jun 24, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
520
Basketball technique action recognition using 3D convolutional neural networks
Jingfei Wang1,2, Liang Zuo3, Carlos Cordente Martínez4
1Physical Education Department, Northwestern Polytechnical University, Xi'an, 710129, Shaanxi, People's Republic of China. jingfei.wang@alumnos.upm.es.
Scientific Reports
|June 7, 2024
Summary
This study introduces a new method using 3D Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks for accurate basketball action recognition. The model significantly improves technique identification for coaches and players.
Area of Science:
- Computer Science
- Artificial Intelligence
- Sports Analytics
Background:
- Automated recognition of sports actions is crucial for performance analysis.
- Basketball technique analysis traditionally relies on manual observation, which is time-consuming and subjective.
Purpose of the Study:
- To develop an accurate and automated system for recognizing basketball technique actions using deep learning.
- To enhance the identification of various actions within basketball games.
Main Methods:
- Implementation of three-dimensional (3D) Convolutional Neural Networks (CNNs) combined with Long Short-Term Memory (LSTM) networks.
- Utilized publicly available basketball action datasets (NTURGB+D, Basketball-Action-Dataset, B3D Dataset) with preprocessing techniques.
- Employed optimization algorithms like adaptive learning rate adjustment and regularization to improve model performance.
Main Results:
- The proposed 3D CNN-LSTM model achieved outstanding performance in basketball technique action recognition.
- Demonstrated significant accuracy improvements: 15.1% over frame difference methods and 12.4% over optical flow methods.
- Showcased strong robustness, achieving an average accuracy of 93.1% across diverse conditions.
Conclusions:
- The developed method effectively captures the spatiotemporal relationships of basketball actions.
- Provides a reliable technical assessment tool for basketball coaches and players.
- Highlights the potential of deep learning for advanced sports analytics.
