Related Experiment Video
Updated: Jun 16, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
496
Research on sports image classification method based on SE-RES-CNN model
Qinglan Li1, Jichong Lei2, Changan Ren3
1Department of Physical Education and Research, Hunan Institute of Technology, Hengyang, 421002, Hunan, China.
Scientific Reports
|August 17, 2024
Summary
A new SE-RES-CNN model enhances sports image classification accuracy by 5% using adaptive channel weighting and deep feature extraction. This deep learning approach significantly improves the efficiency of sports image retrieval and management.
Area of Science:
- Computer Science
- Artificial Intelligence
- Image Processing
Background:
- Manual sports image classification is inefficient and inaccurate for large datasets.
- Advancements in digital technologies necessitate efficient image classification methods.
Purpose of the Study:
- To develop an efficient deep learning model for sports image classification.
- To improve the accuracy and speed of sports image retrieval and management.
Main Methods:
- Introduction of a novel SE-RES-CNN neural network model.
- Integration of an SE module for adaptive channel weight adjustment.
- Utilizing a Res module for deep feature extraction, gradient vanishing prevention, and multi-scale processing.
Main Results:
- The SE-RES-CNN model achieved up to 98% accuracy on a sports image dataset.
- Demonstrated a 5% improvement in classification accuracy compared to VGG-16 and ResNet50.
- Classified 100 images in 6 seconds with a single prediction time of 0.012 seconds.
Conclusions:
- The SE-RES-CNN model is accurate and effective for sports image classification.
- The proposed model significantly enhances the efficiency of sports image retrieval.
- Validated the model's capability in handling large-scale image datasets.
