Skip to main content
. 2021 Apr 20;7:e495. doi: 10.7717/peerj-cs.495

Table 2. A comparison of actual and predicted accuracies of DenseNet121 and InceptionResNetV2.

Pathology Model Actual Predicted
Atelectasis DenseNet121 20.21% 22.10%
InceptionResNetV2 20.51% 21.57%
Cardiomegaly DenseNet121 4.79% 5.71%
InceptionResNetV2 6.15% 5.63%
Consolidation DenseNet121 9.86% 8.56%
InceptionResNetV2 8.30% 8.88%
Edema DenseNet121 4.49% 4.39%
InceptionResNetV2 4.10% 4.67%
Effusion DenseNet121 25.88% 28.07%
InceptionResNetV2 26.07% 25.39%
Emphysema DenseNet121 4.79% 3.59%
InceptionResNetV2 4.49% 4.04%
Fibrosis DenseNet121 3.03% 2.56%
InceptionResNetV2 2.44% 3.26%
Infiltration DenseNet121 40.14% 37.92%
InceptionResNetV2 40.53% 39.67%
Mass DenseNet121 12.30% 11.13%
InceptionResNetV2 13.28% 11.30%
Nodule DenseNet121 11.91% 11.28%
InceptionResNetV2 12.70% 12.21%
Pleural_Thickening DenseNet121 7.23% 6.34%
InceptionResNetV2 7.03% 6.20%
Pneumonia DenseNet121 3.22% 2.56%
InceptionResNetV2 2.83% 2.57%
Pneumothorax DenseNet121 8.89% 8.84%
InceptionResNetV2 9.77% 9.83%