Table 5 Comparison of metrics for evaluating the results of the various network experiments
From: Research and implementation of mural classification based on lightweight network
Network | Params | Flops | Accuracy | Runtime | Frame rate |
|---|---|---|---|---|---|
Resnet18 | 11.69 M | 1.82 G | 87.5% | 0.107 s | 10.31 |
Resnet34 | 21.80 M | 3.67 G | 88.3% | 0.191 s | 5.23 |
RepVGG18 | 11.50 M | 1.80 G | 86.8% | 0.091 s | 10.99 |
ShufflenetV2 | 2.28 M | 0.15 G | 79.8% | 0.096 s | 10.41 |
MoblienetV1 | 5.10 M | 0.57 G | 80.1% | 0.147 s | 6.80 |
MoblienetV2 | 3.50 M | 0.32 G | 81.0% | 0.098 s | 10.20 |
PeleeNet | 2.80 M | 0.52 G | 82.7% | 0.126 s | 7.93 |
Efficientnetv2_b0 | 21.46 M | 2.87 G | 82.1% | 0.129 s | 7.74 |
MobileVit-S | 5.56 M | 1.42 G | 85.5% | 0.51 s | 1.96 |
LightVit-T | 9.4 M | 0.73 G | 89.1% | 0.18 s | 5.55 |
SER-Net | 8.22 M | 0.95 G | 90.1% | 0.09 s | 11.11 |
