Solution for sports image classification using modified MobileNetV3 optimized by modified battle royal optimization
Bing Wang1, Asad Rezaei Sofla2,3
1School of Physical Education, Zhengzhou Normal University, ZhengZhou, HeNan, 450044, China.
Heliyon
|November 29, 2023
Summary
This study introduces a new deep learning framework for sports image classification, using an optimization algorithm to select essential features. The proposed method enhances classification accuracy and reduces data dimensionality for better sports analytics.
Area of Science:
- Computer Vision
- Machine Learning
- Sports Analytics
Background:
- Sports image classification is crucial for analyzing performance and detecting events.
- Current methods may lack efficiency in feature selection and accuracy.
- Automated analysis of sports data offers significant potential for training and strategy.
Purpose of the Study:
- To propose a novel hybrid framework for sports image classification.
- To enhance classification accuracy and reduce image dimensionality using optimization algorithms.
- To demonstrate the effectiveness of the proposed deep learning and optimization approach.
Main Methods:
- Developed a hybrid framework combining deep learning with a modified Battle Royal optimization algorithm.
- Utilized the optimization algorithm as a feature selector to identify essential image features.
- Evaluated the framework on a dataset of sports images.
Main Results:
- The proposed WOA-based framework significantly improved classification accuracy.
- Achieved substantial dimensionality reduction by selecting only essential features.
- Outperformed existing methods in both accuracy and efficiency.
Conclusions:
- The hybrid deep learning and optimization framework is effective for sports image classification.
- Feature selection using optimization algorithms enhances model performance.
- This approach has the potential to advance sports image analysis and applications.
