Table 1 The mean F1 score, precision, recall, and specificity for each deep learning model using 70/30 data split over 10 randomized runs
From: Detection of developmental dysplasia of the hip in X-ray images using deep transfer learning
Deep learning model | F1 score | Precision | Recall | Specificity |
|---|---|---|---|---|
SqueezeNet | 76.6% | 71.0% | 86.4% | 78.7% |
GoogLeNet | 87.8% | 82.3% | 94.4% | 89.1% |
Inceptionv3 | 88.6% | 79.8% | 100.0% | 86.6% |
DenseNet-201 | 85.2% | 75.1% | 98.6% | 83.0% |
MobileNetv2 | 85.3% | 76.7% | 96.7% | 84.3% |
Resnet101 | 85.7% | 76.0% | 98.6% | 83.6% |
Resnet50 | 88.8% | 81.9% | 97.5% | 88.4% |
Resnet18 | 84.7% | 75.3% | 97.2% | 83.0% |
Xception | 84.9% | 74.7% | 98.6% | 82.6% |
Inception-ResNet-v2 | 84.7% | 75.9% | 96.1% | 84.1% |
ShuffleNet | 82.7% | 71.7% | 98.1% | 79.7% |
DarkNet-53 | 88.9% | 80.9% | 98.9% | 87.9% |
EfficientNet-b0 | 78.5% | 65.5% | 98.1% | 73.4% |