Table 8 The comparison of proposed hybrid model with the state-of-the-art methods.
Study | Method | Performance |
|---|---|---|
Deng et al.66 | Customized CNN architecture | ACC: 93.04% |
Abdalla-Aslan et al.67 | segmented using adaptive threshold, Gray level values and shape features, Cubic SVM with Error-Correcting Output Codes for classificatiob | ACC: 93.6% |
Ghaznavi et al.68 | CNN comparison with AlexNet and VGGNet16 | Average PRE: 92% |
Jaiswal et al.69 | Transfer learning (ResNet, MobileNet) + XGBoost | average ACC: 93% |
Rajee and Mythili70 | Inception ResNetV2 | ACC: 94.51% |
Proposed hybrid model | Hybrid approach (MobileNetV2 + Swin Transformer) with bagging ensemble | ACC: 96.5 |
