Table 4.
Data | Method | DeconvNet | Light-Weight |
---|---|---|---|
Freiburg | W/O synthetic data | 73.65(%) | 94.43(%) |
After using two-class synthetic | 66.62(%) | 94.06(%) | |
After using four-class high-definition | 68.7(%) | 94.59(%) | |
After using four-class random synthetic | 68.14(%) | 93.89(%) |