You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I saw these parameters
cmd:option('-content_layers', 'relu4_2', 'layers for content')
cmd:option('-style_layers', 'relu1_1,relu2_1,relu3_1,relu4_1,relu5_1', 'layers for style')
but how do you know which layer in VGG learned style and which learned the content ?
The text was updated successfully, but these errors were encountered:
VGG19 recognizes content features because it has been trained to classify image contents. The lowest layers recognize simple features, such as lines, whereas higher layers see increasingly complex features, such as shapes and objects.
When you select a layer for style, neural-style adds a so-called Gram matrix to this layer to evaluate style.
Does anyone knows why the guys used conv1_1, conv2_1, conv3_1, conv4_1, conv5_1 instead of for example conv1_2, conv2_2, conv3_4, conv4_4, conv5_4 ("last" layer of respective area) in the original paper
I saw these parameters
cmd:option('-content_layers', 'relu4_2', 'layers for content')
cmd:option('-style_layers', 'relu1_1,relu2_1,relu3_1,relu4_1,relu5_1', 'layers for style')
but how do you know which layer in VGG learned style and which learned the content ?
The text was updated successfully, but these errors were encountered: