Table III.
V-Net characteristics used for satisfactory training of 2D and 3D models.
2D | 3D | |
---|---|---|
Architecture | - Number of channels: 16 | - Number of channels: 12 |
- Number of levels: 2 | - Number of levels: 4 | |
- Number of convolutions: [4, 4] | - Number of convolutions: [1, 3, 4, 3] | |
- Bottom convolutions: 2 | - Bottom convolutions: 4 | |
Learning rate | 0.0005 | 0.00001 |
2D, two-dimensional; 3D, three-dimensional.
A convolution is when two sets of information are merged to form a new function.