Table 4 Performance comparison of ResNet34, ResNet50, and ResNet101 under different input modalities
From: A maturity classification model for winter jujubes based on DSAF-ResNet
Inputs | Models | Traning set | Test set | ||||
|---|---|---|---|---|---|---|---|
Accuracy(%) | Precision(%) | Recall(%) | Accuracy(%) | Precision(%) | Recall(%) | ||
GLCM texture features | ResNet34 | 88.87 | 94.72 | 94.71 | 81.22 | 80.90 | 81.22 |
ResNet50 | 91.93 | 96.60 | 96.52 | 79.56 | 80.76 | 79.56 | |
ResNet101 | 84.56 | 88.22 | 84.84 | 79.01 | 83.91 | 79.01 | |
Spectral features | ResNet34 | 89.85 | 94.45 | 93.60 | 86.74 | 89.07 | 86.74 |
ResNet50 | 84.01 | 87.66 | 86.23 | 83.43 | 85.08 | 83.43 | |
ResNet101 | 85.12 | 88.58 | 88.04 | 82.32 | 82.75 | 82.32 | |
Fused spectral-texture features | ResNet34 | 94.44 | 97.61 | 97.50 | 92.27 | 92.80 | 92.27 |
ResNet50 | 93.05 | 95.20 | 95.13 | 91.71 | 92.03 | 91.71 | |
ResNet101 | 93.60 | 94.16 | 93.32 | 89.50 | 90.44 | 89.50 | |
