Table 3 Comparison of CAD-ReconNet and other networks
From: Three-dimensional reconstruction of industrial parts from a single image
Network | Parameters | Maximum training accuracy | Maximum verification accuracy | Maximum testing accuracy |
|---|---|---|---|---|
CAD-ReconNet | Hexagon head bolt | 96.91% | 97.69% | 97.92% |
Cylindrical gear | 94.60% | 91.75% | 89.29% | |
Shoulder ring | 97.50% | 99.34% | 99.70% | |
Hexagon nut | 96.95% | 87.46% | 81.85% | |
Cylindrical roller bearing | 98.49% | 97.36% | 95.24% | |
ResNet-18 | Hexagon head bolt | 94.56% | 94.72% | 92.56% |
Cylindrical gear | 91.25% | 77.89% | 84.52% | |
Shoulder ring | 85.63% | 84.16% | 85.12% | |
Hexagon nut | 86.40% | 73.93% | 78.27% | |
Cylindrical roller bearing | 95.11% | 91.09% | 87.50% | |
ResNet-34 | Hexagon head bolt | 98.31% | 96.70% | 95.54% |
Cylindrical gear | 92.65% | 83.83% | 82.44% | |
Shoulder ring | 98.57% | 92.74% | 92.86% | |
Hexagon nut | 93.01% | 83.83% | 82.14% | |
Cylindrical roller bearing | 95.33% | 93.07% | 87.20% |