Table 4 Results of the evaluation of the MobileNetV2 method on different classes of the applied dataset
From: A novel nondestructive detection approach for seed cotton lint percentage using deep learning
Model | Classes | Accuracy /% | Precision /% | Recall /% | F1-Score /% |
|---|---|---|---|---|---|
MobileNetV2 without transfer learning | A | 97.66 | 92.41 | 95.53 | 93.94 |
B | 96.54 | 91.10 | 92.40 | 91.75 | |
C | 98.81 | 95.31 | 94.95 | 95.13 | |
D | 98.53 | 98.77 | 92.95 | 95.77 | |
E | 97.73 | 92.46 | 93.64 | 93.05 | |
F | 98.26 | 93.82 | 93.51 | 93.66 | |
Average | 97.92 | 93.98 | 93.83 | 93.88 | |
MobileNetV2 with transfer learning | A | 98.78 | 96.63 | 96.98 | 96.80 |
B | 98.58 | 97.53 | 95.59 | 96.55 | |
C | 98.44 | 92.07 | 95.51 | 93.76 | |
D | 98.12 | 95.68 | 93.72 | 94.69 | |
E | 98.48 | 95.93 | 94.56 | 95.44 | |
F | 98.19 | 91.96 | 95.17 | 93.94 | |
Average | 98.43 | 94.97 | 95.26 | 95.20 |