Research Article
Two-Stage Intelligent DarkNet-SqueezeNet Architecture-Based Framework for Multiclass Rice Grain Variety Identification
Table 4
Classification results of the SqueezeNet model (experiment 2).
| Classifier (%) | Recall (%) | Sensitive (%) | F1 score (%) | FNR (%) | Accuracy (%) | Time (sec) |
| Linear SVM | 99.22 | 99.24 | 99.23 | 0.78 | 99.2 | 301.74 | Quadratic SVM | 99.46 | 99.46 | 99.4 | 0.54 | 99.5 | 304.67 | Weight KNN | 98.58 | 98.44 | 98.51 | 1.42 | 98.6 | 362.9 | Cosine KNN | 97.9 | 97.94 | 97.92 | 2.1 | 97.9 | 387.4 | Linear discriminant | 99.32 | 99.32 | 99.32 | 0.68 | 99.3 | 72.783 | Medium neural network | 99.42 | 99.42 | 99.42 | 0.58 | 99.4 | 178.13 | Narrow neural network | 99.38 | 99.38 | 99.38 | 0.62 | 99.4 | 244.41 | Wide neural network | 99.41 | 99.4 | 99.41 | 0.58 | 99.4 | 249.99 | Bilayered neural network | 99.36 | 99.36 | 99.36 | 0.64 | 99.4 | 336.25 | Trilayered neural network | 99.37 | 99.38 | 99.37 | 0.64 | 99.4 | 458.32 |
|
|