TABLE 7.
Extended experiments on the application of progressive learning strategies to 3D U-Net.
| Method | PROMISE 2012 dataset | |
|---|---|---|
| IoU (%) | Dice (%) | |
| 3D U-Net Çiçek et al. (2016) | 56.84 | 72.48 | 
| 3D U-Net + IPL | 57.58 | 73.08 | 
| 3D U-Net + IPL + EPL | 58.53 | 73.84 | 
The bold values indicates the best performance.