Fig 1.
Flowchart of deep models for transfer learning fine-tuning.
Fig 2.
Flowchart overview of the proposed method.
(a) Input raw image dataset. (b) Data preprocessing (c) Pre-trained deep learning models and extract bottleneck features. (c-1) Example of basic network architecture (VGG16). (c-2) Five basic deep learning models (d) Classify with machine learning classifier. (d-1) Five traditional machine learning classification methods.
Fig 3.
X-ray images dataset for normal cases (first row) and COVID-19 patients (second row).
Table 1.
The division of training set and test set.
Table 2.
Images sample source.
Table 3.
Evaluation results of transfer learning methods.
Table 4.
Evaluation results of VGG16 combined with different classifiers.
Table 5.
Evaluation results of InceptionV3 combined with different classifiers.
Table 6.
Evaluation results of ResNet50 combined with different classifiers.
Table 7.
Evaluation results of DenseNet121 combined with different classifiers.
Table 8.
Evaluation results of Xception combined with different classifiers.
Table 9.
Performance of the best method on other dataset [16].
Table 10.
Comparison of results between our proposed method and other methods.