use valid padding in first conv layer of synthesis generator

bodokaiser / mrtoct-tensorflow

Tensorflow models for MRI to CT synthesis.

https://arxiv.org/abs/1901.05259

26 stars 5 forks source link

use valid padding in first conv layer of synthesis generator #10

Open bodokaiser opened 7 years ago

bodokaiser commented 7 years ago

If we use padding='same' for the first convolution layer with kernel_size=9 we get output shape [batch_size, 16, 16, 16, 1] as intended from the original paper.