表 3. Classification performance of different backbone networks for IIMs.
不同主干网络对IIMs的分类效果
| 网络名称 | 正确率 | 精确度 | 灵敏度 | 综合指标 | 参数量 | 计算量 |
| 注：加粗数字为最优结果 | ||||||
| GhostNet | 86.5% | 85.9% | 84.5% | 85.1% | 3.9 MB | 144.3 MB |
| EfficientNetV1 | 80.2% | 79.2% | 76.9% | 77.9% | 4.1 MB | 396.9 MB |
| EfficientNetV2 | 84.2% | 82.9% | 80.7% | 81.5% | 7.1 MB | 726.5 MB |
| MobileNetV1 | 91.6% | 91.4% | 90.2% | 90.7% | 4.9 MB | 574.5 MB |
| MobileNetV2 | 91.0% | 91.1% | 90.0% | 90.5% | 2.2 MB | 323.2 MB |
| MobileNetV3_small | 85.5% | 85.5% | 82.7% | 83.7% | 1.7 MB | 60.0 MB |
| MobileNetV3_large | 90.4% | 89.6% | 88.9% | 89.2% | 4.2 MB | 222.9 MB |
| 主干网络（LCNN） | 93.5% | 93.3% | 92.5% | 92.8% | 0.5 MB | 82.1 MB |