摘要
针对图像分类学习不够深入的问题,提出图像分类问题的几种深度学习策略研究。通过分析当前主流的主动深度学习图像、多标签图像和多尺度网络图像三种深度学习方法的工作原理和存在的优势与不足,探讨图像分类问题的优化学习策略。随后采用图像分类问题的几种深度学习策略实验的方式加以对比,实验结果表明,参数共享的深度学习图像分类方法不仅提高了预测速度,而且还能确保模型的准确性。
Aiming at the problem that image classification learning is not deep enough,several deep learning strategies for image classification are proposed.By analyzing the working principle,advantages and disadvantages of three mainstream deep learning methods:active deep learning image,multi label image and multi-scale network image,this paper discusses the optimal learning strategy of image classification.Then,several deep learning strategies for image classification are compared.The experimental results show that the parameter sharing deep learning image classification method not only improves the prediction speed,but also ensures the accuracy of the model.
作者
李建伟
LI Jianwei(School of Electronic Science and technology,Department of Information Science,Beijing University of Technology,Beijing 100124,China)
出处
《数字通信世界》
2022年第1期67-69,共3页
Digital Communication World
关键词
图像分类
分类方法
深度学习
学习策略
image classification
classification method
deep learning
learning strategy
