Table 1.
Initial Learning Rate | Decreased Epochs | Decreased Ratio | Max Epoch | Optimizer | Weight Decay |
---|---|---|---|---|---|
0.01 | [50, 70, 80] | 0.4 | 100 | Adam | 0 |
3D RPN and 3DDLNN are the CNNs borrowed from Liao et al. (Liao et al., 2019) to extract scan-level feature.