Related Experiment Video
Updated: Jun 28, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
529
The application of improved densenet algorithm in accurate image recognition
Yuntao Hou1, Zequan Wu2, Xiaohua Cai2
1Heilongjiang Academy of Agricultural Machinery Sciences, Heilongjiang Academy of Agricultural Sciences, Harbin, 150081, China. legend@haas.cn.
Scientific Reports
|April 15, 2024
Summary
This study enhances image recognition using improved dense convolutional networks and gradient quantization for efficient parallel processing. The optimized model achieves higher accuracy and faster training speeds in computer vision applications.
Area of Science:
- Artificial Intelligence
- Computer Vision
- Machine Learning
Background:
- Image recognition is crucial in artificial intelligence and computer vision.
- Existing methods face technical challenges in feature reuse and computational efficiency.
Purpose of the Study:
- To improve feature reuse in dense convolutional networks for enhanced image recognition.
- To optimize parallel algorithms for faster and more efficient model training.
Main Methods:
- Improved feature reuse in dense convolutional networks.
- Gradient quantization applied to traditional parallel algorithms.
- Layer-by-layer independent parameter updates.
- Introduction of quantization error to mitigate gradient loss.
Main Results:
- Enhanced model parameter efficiency while maintaining recognition accuracy.
- Demonstrated superior performance compared to Visual Geometry Group and EfficientNet models.
- Optimized parallel algorithm significantly improves acceleration ratio and training speed.
- Reduced communication time and data volume in parallel processing.
Conclusions:
- The proposed strategies improve the accuracy and training speed of image recognition technology.
- The optimized model expands the application of image recognition in computer vision.
- Gradient quantization offers a better alternative to traditional parallel algorithms for image recognition tasks.
