Table 3. Comparison of predicted scores and ground truth scores for various sets of images from the LaMem dataset.
Comparison table between different networks with the proposed one.
| Dataset | MemNet | MCDRNet | EMNet | ResMem-Net | Ground Truth |
|---|---|---|---|---|---|
| Top 10 | 91.70% | 93.15% | 91.89% | 93.82% | 100% |
| Top 25 | 90.40% | 91.76% | 91.36% | 92.8% | 100% |
| Top 50 | 89.57% | 91.27% | 90.85% | 90.52% | 99.35% |
| Top 100 | 89.17% | 90.42% | 90.38% | 90.74% | 98.45% |
| Top 200 | 88.91% | 90.00% | 90.02% | 90.65% | 97.57% |
| Bottom 200 | 55.06% | 54.57% | 54.23% | 52.18% | 42.16% |
| Bottom 100 | 54.35% | 52.79% | 52.74% | 49.13% | 39.01% |
| Bottom 50 | 54.20% | 51.65% | 51.25% | 48.72% | 36.3% |
| Bottom 25 | 54.44% | 51.20% | 50.34% | 48.27% | 34.41% |
| Bottom 10 | 58.06% | 50.94% | 48.41% | 47.9% | 33.57% |