Table 7 Comparison of different parameters before and after compression on the faster-CNN model
Evaluation parameter | Without pruning | 50–80% pruning | 60–90% pruning | 70–90% pruning | 80–90% pruning |
|---|---|---|---|---|---|
Accuracy | 96.92% | 96.85% | 96.03% | 92.35% | 80.36% |
Loss | 0.1817 | 0.1328 | 0.1557% | 0.2067% | 0.50% |
Precision | 96.07% | 94.85% | 96.05% | 93.08% | 86.41% |
Recall | 95.5% | 94.85% | 94.13% | 92.35% | 69.45% |
F-score | 95.78% | 94.85% | 92.71% | 92.71% | 77% |
Size | 53.78 MB | 50.91 MB | 28.16 MB | 22.81 MB | 17.33 MB |
Pruning + post-quantization | |||||
Size | 14.11 MB | 8.21 MB | 8.23 MB | 6.62 MB | 4.81 MB |
Accuracy | 89.2% | 83.3% | 87.2% | 78.8% | 79.2% |