def get_gradient_kernel(self):
gradient_block = nn.ModuleList()
conv_x = Conv2d(1, 1, [1, 2], padding='SAME',
bias=False, weight=[[[[-1, 1]]]])
gradient_block.append(conv_x)
conv_y = Conv2d(1, 1, [2, 1], padding='SAME',
bias=False, weight=[[[[-1], [1]]]])
gradient_block.append(conv_y)
return gradient_block
def get_centered_gradient_kernel(self):
centered_gradient_block = nn.ModuleList()
conv_x = Conv2d(1, 1, [1, 3], padding='SAME',
bias=False, weight=[[[[-0.5, 0, 0.5]]]])
centered_gradient_block.append(conv_x)
conv_y = Conv2d(1, 1, [3, 1], padding='SAME',
bias=False, weight=[[[[-0.5], [0], [0.5]]]])
centered_gradient_block.append(conv_y)
return centered_gradient_block
def get_divergence_kernel(self):
divergence_block = nn.ModuleList() #[conv_x, conv_y]
conv_x = Conv2d(1, 1, [1, 2], padding='SAME',
bias=False, weight=[[[[-1, 1]]]])
divergence_block.append(conv_x)
conv_y = Conv2d(1, 1, [2, 1], padding='SAME',
bias=False, weight=[[[[-1], [1]]]])
divergence_block.append(conv_y)
return divergence_block
def get_gray_conv(self):
gray_conv = Conv2d(3,
1, [1, 1],
bias=False,
padding='SAME',
weight=[[[[0.114]], [[0.587]], [[0.299]]]])
return gray_conv
def get_gaussian_conv(self):
gaussian_conv = Conv2d(1, 1, [5, 5], bias=False, padding='SAME',
weight=[[[[0.000874, 0.006976, 0.01386, 0.006976, 0.000874],
[0.006976, 0.0557, 0.110656, 0.0557, 0.006976],
[0.01386, 0.110656, 0.219833, 0.110656, 0.01386],
[0.006976, 0.0557, 0.110656, 0.0557, 0.006976],
[0.000874, 0.006976, 0.01386, 0.006976, 0.000874]]]])
return gaussian_conv