全文下载: 202203015.pdf
文章编号: 1672-6987(2022)03010705; DOI: 10.16351/j.1672-6987.2022.03.015
王剑峰,王传旭(青岛科技大学 信息科学技术学院,山东 青岛 266061)
摘要: 提出了一种基于DenseNet模型优化的新冠肺炎CT图像检测算法。首先,分析并实现了ResNet、EfficientNet和DenseNet 3种网络模型,完成了在FlyAI平台上对新冠肺炎CT图像检测训练及分类任务。通过对技术原理、分类准确率等参量的分析和对比,表明DenseNet具有优越的自适应能力和分类能力。进一步针对该模型改用Focal Loss损失函数模型,在新冠肺炎CT图像数据集上的识别和分类达到9457%的准确率,相比原来交叉熵损失函数下的模型,提升226%的精度,同时也高于其他检测方法,证明了所提算法的有效性。
关键词: 新冠肺炎CT图像; ResNet模型; EfficientNet模型; DenseNet模型
中图分类号: TP 301.6文献标志码: A
引用格式: 王剑峰, 王传旭. 基于DenseNet模型优化的新冠肺炎CT图像检测算法[J]. 青岛科技大学学报(自然科学版), 2022, 43(3): 107111.
WANG Jianfeng, WANG Chuanxu. COVID19 CT Image detection based on optimized denseNet model[J]. Journal of Qingdao University of Science and Technology(Natural Science Edition), 2022, 43(3): 107111.
COVID19 CT Image Detection Based on Optimized DenseNet Model
WANG Jianfeng, WANG Chuanxu
(College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China)
Abstract: This paper proposes a detection algorithm for COVID19 CT image based on DenseNet model optimization. Firstly, three models are analyzed and implemented, which are ResNet, EfficientNet and DenseNet, tested on FlyAI platform. Through the analysis and comparison of their technical principles, classification accuracy and other parameters, we find that DenseNet has superior adaptive and classification capabilities. Secondly, the focus loss function model is used to improve this model. Its recognition and classification accuracy on the CT image data set of COVID19 reaches 9457%, which increased by 226% compared with the original model under the cross entropy loss function. At the same time, it is also higher than other detection methods, which proves the effectiveness of the proposed algorithm.
Key words: COVID19 CT; deep learning; ResNet; EfficientNet; DenseNet
基金项目: 国家自然科学基金项目(61672305).
作者简介: 王剑峰 (1980—), 男,收稿日期: 20210706
实验师.