新冠肺炎胸部CT检测的基准深度学习模型及自动化模型设计

June 10, 2020

2020   ·   techniques     ·   techniques  

原文: http://zhuanlan.zhihu.com/p/147361922

2. Experimental Results

2.1 Benchmark Deep Learning Models

If you want to run the benchmark experiments, you can refer to the directory of covid19_pipeline.

  • The pipeline of benchmarking deep learning-based models.

pipeline

  • Performance comparison between different models

model_perf

  • Performance comparison between ResNet3d models with different depth

model_depth

  • Performance comparison between models trained by scan data comprising a different number of slices.

mdoel_slice

  • The model accuracy before and after using MixUp data augmentation method.

model_mixup

2.2 Automated Model Design (AutoML)

The code of NAS will be released very soon …

  • NAS pipeline

NAS pipeline

  • Search space

Search space

  • The performance comparison between baseline models and models designed by NAS

nas_vs_manual

3. Citation

Benchmarking Deep Learning Models and Automated Model Design for COVID-19 Detection with Chest CT Scans

@article {He et al.benchmark,
	author = {He, Xin and Wang, Shihao and Shi, Shaohuai and Chu, Xiaowen and Tang, Jiangping and Liu, Xin and Yan, Chenggang and Zhang, Jiyong and Ding, Guiguang},
	title = {Benchmarking Deep Learning Models and Automated Model Design for COVID-19 Detection with Chest CT Scans},
	elocation-id = {2020.06.08.20125963},
	year = {2020},
	doi = {10.1101/2020.06.08.20125963},
	publisher = {Cold Spring Harbor Laboratory Press},
	URL = {https://www.medrxiv.org/content/early/2020/06/09/2020.06.08.20125963},
	eprint = {https://www.medrxiv.org/content/early/2020/06/09/2020.06.08.20125963.full.pdf},
	journal = {medRxiv}
}

微信公众号：AutoML机器学习

MARSGGBO♥原创
如有意合作或学术讨论欢迎私戳联系~
邮箱:marsggbo@foxmail.com
2020-06-08 11:56:22

Flag Counter