Table 7 Performance comparison of proposed hMPV-net and existing DL models.
From: Deep learning approach for automated hMPV classification
Metric | ResNet-50 | VGG-16 | Proposed hMPV-Net |
|---|---|---|---|
Dataset Size | 5000 images | 5000 images | 10,000 images |
Training Accuracy | 85.60% | 85.60% | 99.20% |
Validation Accuracy | 78.30% | 78.30% | 93.50% |
Test Accuracy | 81.40% | 81.40% | 91.80% |
Loss Value | 0.62 | 0.62 | 0.2154 |
Computational Cost (GFLOPs) | 15.6 GFLOPs | 15.6 GFLOPs | 3.2 GFLOPs |
ROC-AUC Score | 0.85 | 0.85 | 0.91 |
Key Features | Deep residual learning | Deep convolutional layers | Lightweight, pathogen-specific, class-weighted loss, data augmentation |
