Table 2.
Classifier | Full Spectra (%) | Effective Wavelengths (%) | ||||
---|---|---|---|---|---|---|
Calibration | Validation | Prediction | Calibration | Validation | Prediction | |
CNN-SoftMax a | 91.191 | 89.065 | 88.838 | 87.629 | 84.071 | 82.860 |
CNN-LR | 94.060 | 88.611 | 87.752 | 90.070 | 83.731 | 83.276 |
CNN-PLS-DA | 91.112 | 88.082 | 86.644 | 87.088 | 82.709 | 82.027 |
CNN-SVM | 93.695 | 89.255 | 88.006 | 89.970 | 84.487 | 84.260 |
ResNet-SoftMax | 95.381 | 85.698 | 86.039 | 92.273 | 79.985 | 79.228 |
ResNet-LR | 99.585 | 84.335 | 82.324 | 98.238 | 76.040 | 75.952 |
ResNet-PLS-DA | 95.130 | 85.585 | 85.358 | 91.707 | 78.509 | 77.677 |
ResNet-SVM | 96.325 | 85.963 | 85.887 | 94.098 | 79.153 | 79.115 |
LR | 84.156 | 82.406 | 83.012 | 62.736 | 62.429 | 65.305 |
PLS-DA | 81.764 | 79.947 | 80.401 | 78.870 | 77.261 | 77.147 |
SVM | 93.557 | 89.217 | 88.422 | 89.441 | 84.147 | 84.033 |
a. CNN-SoftMax means using SoftMax function as classifier for the CNN model.