def __init__(self, in_planes=17, out_planes=8):
super(SegRefineNet, self).__init__()
self.conv1 = nn.Sequential(
conv2d_lrelu(in_planes, out_planes, kernel_size=3, stride=1,
pad=1))
self.classif1 = nn.Conv2d(out_planes,
1,
kernel_size=3,
padding=1,
stride=1,
bias=False)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2.0 / n))
elif isinstance(m, nn.Conv3d):
n = m.kernel_size[0] * m.kernel_size[1] * m.kernel_size[
2] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2.0 / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.BatchNorm3d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
m.bias.data.zero_()
def __init__(self, inplanes, planes, stride, downsample, pad, dilation):
super(BasicBlock, self).__init__()
self.conv1 = conv2d_lrelu(inplanes, planes, 3, stride, pad, dilation)
self.conv2 = conv2d(planes, planes, 3, 1, pad, dilation)
self.downsample = downsample
self.stride = stride
def __init__(self, out_planes=32):
super(DispRefineNet, self).__init__()
self.out_planes = out_planes
self.conv2d_feature = conv2d_lrelu(in_planes=4,
out_planes=self.out_planes,
kernel_size=3,
stride=1,
pad=1,
dilation=1)
self.residual_astrous_blocks = nn.ModuleList()
astrous_list = [1, 2, 4, 8, 1, 1]
for di in astrous_list:
self.residual_astrous_blocks.append(
BasicBlock(self.out_planes,
self.out_planes,
stride=1,
downsample=None,
pad=1,
dilation=di))
self.conv2d_out = nn.Conv2d(self.out_planes,
1,
kernel_size=3,
stride=1,
padding=1)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2.0 / n))
elif isinstance(m, nn.Conv3d):
n = m.kernel_size[0] * m.kernel_size[1] * m.kernel_size[
2] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2.0 / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.BatchNorm3d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
m.bias.data.zero_()
return