Table 2.
A comparative performance of the employed models. Average computational times for each encoder of 35 results during training and validation phases
| Encoder | Size (MB) | Training time (s) | Prediction time (s) | Parameters | Layers |
|---|---|---|---|---|---|
| U-NET | 30 | 381 | 1.1 | 7,760,642 | 39 |
| Modified U-NET | 30 | 385 | 1.3 | 7,763,050 | 74 |
| VGGNet | 71 | 540 | 1.6 | 18,540,938 | 56 |
| ResNet | 118 | 446 | 2.3 | 30,546,458 | 223 |
| DenseNet | 51 | 482 | 3.2 | 12,947,674 | 474 |
| Xception | 103 | 580 | 1.9 | 26,769,602 | 184 |
| MobileNet | 30 | 385 | 1.5 | 7,590,746 | 129 |
| NASNet | 37 | 684 | 4.4 | 8,652,846 | 818 |
| MobileNetV2 | 22 | 386 | 1.8 | 5,591,770 | 202 |
Best values are shown in bold