Table 2.
Evaluation results for the CIFAR-10 data set grouped by network sizes in FLOPs. The first group contains larger configurations, while the second group comprises smaller ones.
| Model | Widening Factor k | Mean Acc | Mil. FLOPs |
|---|---|---|---|
| EffNet V1 large | 0.99 | 85.02% | 79.8 |
| MobileNet large | 0.14 | 78.18% | 11.6 |
| ShuffleNet large | 0.14 | 77.90% | 11.1 |
| EffNet V1 | 0.14 | 80.20% | 11.4 |
| EffNet V2 | 0.22 | 81.67% | 18.1 |
| MobileNetV2 | 0.20 | 76.47% | 16.4 |
| IoTNet-3-4 | 0.7 | 89.9% | 9.9 |
| MobileNet | 0.07 | 77.48% | 5.8 |
| ShuffleNet | 0.06 | 77.3% | 4.7 |
| IoTNet-3-2 | 0.68 | 87.19% | 4.2 |