5,337
edits
Line 21: | Line 21: | ||
It consists of 9 convolution blocks, one for each resolution from <math>4^2</math> to <math>1024^2</math>.<br> | It consists of 9 convolution blocks, one for each resolution from <math>4^2</math> to <math>1024^2</math>.<br> | ||
Each block consists of upsample, 3x3 convolution, AdaIN, 3x3 convolution, AdaIN. | Each block consists of upsample, 3x3 convolution, AdaIN, 3x3 convolution, AdaIN. | ||
After each convolution layer, a gaussian noise with learned variance is added to the feature maps. | After each convolution layer, a gaussian noise with learned variance (block B in the figure) is added to the feature maps. | ||
====Adaptive Instance Normalization==== | ====Adaptive Instance Normalization==== | ||
Each AdaIN block takes as input the latent style <math>w</math> and the feature map <math>x</math>.<br> | |||
An affine layer (fully connected with no activation, block A in the figure) converts the style to a mean <math>y_{b,i}</math> and standard deviation <math>y_{s,i}</math>.<br> | |||
Then the feature map is shifted and scaled to have this mean and standard deviation.<br> | |||
* <math>\operatorname{AdaIN(\mathbf{x}_i, \mathbf{y}) = \mathbf{y}_{s,i}\frac{\mathbf{x}_i - \mu(\mathbf{x}_i)}{\sigma(\mathbf{x}_i)} + \mathbf{y}_{b,i}</math> | |||
==Results== | ==Results== |