Table 4.
Summary of performance metrics of Fet-Net during the ablation study.
| Component(s) Removed (Sequentially) | Average Accuracy (%) | Average Loss | Number of Parameters |
|---|---|---|---|
| Full Architecture | 97.68 | 0.06828 | 10,556,420 |
| Dropout in Feature Extraction Section | 96.58 | 0.1614 | 10,561,028 |
| Dropout in Feature Extraction and Classification Sections | 94.97 | 0.30464 | 10,561,028 |
| Dense Layer with 256 Neurons | 94.51 | 0.28262 | 4,237,572 |
| Second Convolutional Layer with 512 Filters | 91.70 | 0.54048 | 2,238,212 |
| First Convolutional Layer with 512 Filters | 88.38 | 0.92238 | 1,518,852 |
| Second Convolutional Layer with 256 Filters | 87.11 | 0.75504 | 3,693,572 |
| First Convolutional Layer with 256 Filters (1 filter left for functional purposes) | 78.84 | 1.11718 | 14,432 |