Related Experiment Video
Updated: Jun 23, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
517
Improved sports image classification using deep neural network and novel tuna swarm optimization
Zetian Zhou1, Heqing Zhang2, Mehdi Effatparvar3,4
1School of Physical Education and Sports Science, South China Normal University, Guangzhou, 510631, Guangdong, China.
Scientific Reports
|June 19, 2024
Summary
This study introduces a novel deep neural network (DNN) optimized with novel tuna swarm optimization (NTSO) for superior sports image classification. The DNN/NTSO model achieves high accuracy, outperforming existing methods in real-world sports scenarios.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Sports image classification is challenging due to variations in lighting, background, and motion.
- Accurate classification is crucial for sports analytics, media, and broadcasting.
Purpose of the Study:
- To develop a highly accurate and robust model for sports image classification.
- To optimize deep neural network (DNN) hyperparameters using a metaheuristic algorithm.
Main Methods:
- A novel approach combining a deep neural network (DNN) with novel tuna swarm optimization (NTSO) for hyperparameter tuning.
- Rigorous experimentation using a fivefold cross-validation technique on an extensive sports image dataset.
Main Results:
- The DNN/NTSO model achieved high precision (97.665%), recall (95.400%), and F1-score (0.8787).
- Demonstrated superior performance compared to state-of-the-art methods like AGTH-Net, PSO, and YOLOv5.
- Validated effectiveness in real-world scenarios with dynamic conditions and various sports categories.
Conclusions:
- The DNN/NTSO model offers a robust and efficient solution for sports image classification.
- Its performance and scalability make it suitable for real-time sports analytics and media platforms.
