Table 3.
Performance table of training models.
| Transferred models | Accuracy | GPU time (s) | Parameters | Processing time (s) |
|---|---|---|---|---|
| MoblieNet-V2 | 93.455 | 27,240 | 2,235,200 | 0.0374 |
| MoblieNet-V3 | 93.884 | 24,758 | 2,946,622 | 0.0357 |
| Inception-V4 | 93.000 | 98,270 | 42,681,353 | 0.1309 |
| ResNet50 | 93.581 | 51,098 | 25,557,032 | 0.0668 |
| ResNet101 | 93.632 | 78,844 | 42,516,552 | 0.1099 |
| Inception-ResNet-V2 | 94.617 | 111,849 | 54,318,760 | 0.1604 |
| DensNet-BC121 | 94.188 | 54,192 | 6,962,056 | 0.0859 |
| DensNet-BC161 | 94.564 | 78,453 | 26,489,672 | 0.1707 |
| DensNet-BC169 | 94.541 | 56,477 | 12,497,800 | 0.1090 |
| DensenetBC1215 | 94.364 | 56,079 | 7,548,920 | 0.0809 |
| DensenetBC1615 | 95.099 | 80,895 | 27,893,456 | 0.4512 |
| DensenetBC1695 | 94.339 | 58,209 | 13,122,040 | 0.1318 |
| Ensemble | – | – | – | 0.5708 |
GPU time is the processing power needed for training and validation the model. Processing time means the time of each model to identify the same input image.