Table 3 Ablation experiments performed to demonstrate the effectiveness of the two attention modules.
From: Multi-axis compression fusion network for vehicle re-identification
Method | mAP | Rank-1 | Rank-5 |
|---|---|---|---|
Baseline | 76.2% | 92.6% | 97.9% |
Baseline+S-MCF | 78.2% | 93.5% | 98.3% |
Baseline+C-MCF | 78.5% | 94.0% | 98.4% |
Baseline+S-MCF+C-MCF | 82.1% | 95.3% | 98.8% |
MCF-Net (w/o S-MCF) | 82.5% | 95.2% | 98.7% |
MCF-Net (w/o C-MCF) | 82.2% | 95.3% | 98.7% |
MCF-Net (w/o S-MCF+C-MCF) | 81.3% | 95.2% | 98.5% |
MCF-Net (Ours) | 83.9% | 96.9% | 99.0% |
