TABLE 6.
Performance analysis of proposed method with recent techniques.
| Method | Total trainable model parameters (Million) | Precision | Recall | F1-score | Accuracy |
| Residual Net (Sravan et al., 2021) | 22 | 99.28% | 99.26% | 99.27% | 99.26% |
| L-CSMS (Xiang et al., 2021) | 5.44 | – | – | – | 97.90% |
| GoogleNet (Mohanty et al., 2016) | 5 | 99.35% | 99.35% | 99.35% | 99.35% |
| Custom CNN(Geetharamani and Pandian, 2019) | _ | 96.47% | 99.89% | 98.15% | 96.46% |
| MobileNet-Beta (Chen et al., 2020b) | _ | – | 99.01% | – | 99.85% |
| Proposed | 14.4 | 99.63% | 99.93% | 99.78% | 99.99% |
Bold means the architectures are improved.