Table 5 Results of pruning ResNet-50 on ILSVRC-12
From: An efficient pruning scheme of deep neural networks for Internet of Things applications
Method | Top-1 | Top-1 | Top-5 | Top-5 | FLOPs | Parameters |
|---|---|---|---|---|---|---|
| Â | baseline (%) | Accuracy (%) | baseline (%) | Accuracy (%) | pruned (%) | pruned (%) |
CP[31] | - | - | 92.20 | ↓ 1.40 | 50.0 | - |
ThiNet[20] | 72.88 | ↓ 1.87 | 91.14 | ↓ 1.12 | 55.8 | 51.6 |
SFP[37] | 76.15 | ↓ 1.54 | 92.87 | ↓ 0.81 | 41.8 | - |
CFP[32] | - | - | 92.20 | ↓ 0.80 | 49.6 | - |
DCP[30] | 76.01 | ↓ 1.06 | 92.93 | ↓ 0.61 | 55.7 | 51.5 |
FPGM[45] | 76.15 | ↓ 1.21 | 92.87 | ↓ 0.48 | 42.2 | - |
PFS[38] | 77.20 | ↓ 1.60 | - | - | 51.2 | 57.2 |
GAL[35] | 76.15 | ↓ 4.20 | 92.87 | ↓ 1.93 | 43.0 | 16.9 |
ASS[39] | 76.01 | ↓ 2.49 | 92.96 | ↓ 1.45 | 56.6 | 56.0 |
Ours | 76.13 | ↓ 2.38 | 92.86 | ↓ 1.09 | 49.5 | 66.3 |