Related Experiment Video
Updated: Jan 9, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
725
RETRACTED: Human swimming posture recognition combining improved 3D convolutional network and attention residual
1Institute of Physical Education, Yancheng Institute of Technology, Yancheng, China.
Plos One
|December 2, 2025
Summary
This study introduces a novel human swimming posture recognition model using a 3D convolutional neural network. The model achieves high accuracy (95%) and efficiency, crucial for improving training and reducing injuries.
Area of Science:
- Sports Science
- Computer Vision
- Biomechanical Analysis
Background:
- Accurate human swimming posture recognition is vital for enhancing training and preventing sports injuries.
- Existing methods struggle with precise recognition in underwater environments.
Purpose of the Study:
- To develop a novel human swimming posture recognition model with improved accuracy and efficiency.
- To address the limitations of current techniques in underwater swimming analysis.
Main Methods:
- Utilized a 3D convolutional neural network as the foundational model.
- Incorporated global average pooling and batch normalization for optimization.
- Integrated full pre-activation residual network and a three-branch convolutional attention mechanism for enhanced feature extraction.
Main Results:
- Achieved a highest recognition accuracy of 95%, recall of 93.26%, and F1 score of 92.87%.
- Demonstrated low pose recognition errors for freestyle (4.7%), breaststroke (4.9%), butterfly (2.1%), and backstroke (6.6%).
- Achieved the shortest recognition time of 6.78 seconds for freestyle, outperforming existing models.
Conclusions:
- The proposed model offers significant advantages in recognition accuracy and computational efficiency.
- This advancement provides effective support for analyzing athletes' swimming postures.
- The model has the potential to revolutionize swimming training and injury prevention strategies.
