Table 1.
The training time column lists the average seconds/step over 15 epochs, with batch size of eight. The detecting time column denotes the seconds/step for predicting on 60 images, with prediction batch size of one. The total epochs column gives the total epochs of training over the 10 folds from cross-validation.
| Training Time: Average Seconds / Step | Detecting Time | Total Epochs | |
|---|---|---|---|
|
ResNet
|
5.372 seconds | 1.015 seconds | 2808 |
|
DenseNet
|
6.539 seconds | 1.020 seconds | 3315 |
|
EfficientNet
|
6.413 seconds | 1.014 seconds | 3137 |
|
Inception Module
Network
|
1.037 seconds | 0.004 seconds | 2080 |
|
PhiNet
|
1.039 seconds | 0.005 seconds | 2652 |
| Ablated PhiNet | 0.087 seconds | 0.004 seconds | 2748 |