Table 3 Comparison of the performance of different models.
Model | Accuracy/% | Precision/ % | Recall/ % | F1-score/ % | FLOPs | Parameters | Inference time/ms |
|---|---|---|---|---|---|---|---|
EfficientNet_b0 | 87.33 ± 0.68 | 87.61 ± 0.75 | 86.94 ± 0.76 | 87.11 ± 0.78 | 4.11 × 108 | 4.01 × 106 | 54.85 |
Inception_v3 | 88.00 ± 0.95 | 88.18 ± 1.05 | 87.76 ± 1.12 | 87.80 ± 0.97 | 2.86 × 109 | 2.38 × 107 | 108.64 |
MobileNet_v2 | 84.50 ± 0.87 | 85.53 ± 0.82 | 84.12 ± 0.78 | 84.33 ± 0.79 | 3.27 × 108 | 3.50 × 106 | 40.62 |
MobileNetV3_small_100 | 72.40 ± 1.25 | 74.34 ± 1.22 | 72.16 ± 1.17 | 72.42 ± 1.28 | 6.12 × 107 | 1.52 × 106 | 35.18 |
ShuffleNet V2 1 × | 83.82 ± 0.42 | 84.49 ± 0.58 | 82.30 ± 0.54 | 83.45 ± 0.58 | 1.52 × 108 | 2.28 × 106 | 38.83 |
ResNext50_32 × 4d | 84.95 ± 0.62 | 85.13 ± 0.71 | 84.84 ± 0.68 | 84.75 ± 0.57 | 4.29 × 109 | 2.50 × 107 | 107.45 |
ResNet50 | 70.81 ± 0.54 | 72.70 ± 0.62 | 71.05 ± 0.45 | 71.06 ± 0.57 | 4.13 × 109 | 2.55 × 107 | 83.12 |
Xception | 91.18 ± 1.05 | 91.32 ± 0.92 | 90.90 ± 1.11 | 90.98 ± 0.94 | 4.60 × 109 | 2.08 × 107 | 101.25 |
DensNet121 | 82.69 ± 0.87 | 83.76 ± 0.78 | 82.50 ± 0.92 | 82.80 ± 0.86 | 2.90 × 109 | 7.98 × 106 | 116.86 |
ShuffleNet_OD_CA | 91.18 ± 0.45 | 91.49 ± 0.56 | 91.04 ± 0.58 | 91.14 ± 0.53 | 8.03 × 107 | 1.82 × 106 | 36.96 |
