Table 7 Comparative evaluation with SOTA (state-of-the-art models)
From: Vi-GeN 1.0: GAN-Augmented Vision Transformer Pipeline for Diversified Pneumonia Classification
Method | Model type | Dataset used | Accuracy (%) | Sensitivity (%) | Specificity (%) | AUC-ROC |
|---|---|---|---|---|---|---|
CheXNet [Rajpurkar et al.] [20] | DenseNet-121 | ChestX-ray14 | 76.5 | 74.7 | 78.1 | 0.841 |
ResNet-50 + Data Aug [Dash et al.] [6] | CNN + traditional Aug | Kaggle CXR pneumonia | 91.2 | 89.5 | 92.8 | 0.936 |
CoVaD-GAN + ResNet [Dash et al.] [21] | GAN + CNN | COVIDx | 94.5 | 95.0 | 93.7 | 0.951 |
xViTCOS [Mondal et al. [7] | Vision Transformer | COVIDx | 96.2 | 96.0 | 95.8 | 0.963 |
Vi-GeN 1.0 (Ours) | ViT + WGAN-GP | Kaggle CXR pneumonia (balanced) | 97.3 | 98.1 | 96.5 | 0.985 |