def __init__(self, in_planes, planes, stride=1):
super(PreActBlock, self).__init__()
self.sn1 = sn.SwitchNorm2d(in_planes)
self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.sn2 = sn.SwitchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
if stride != 1 or in_planes != self.expansion*planes:
self.shortcut = nn.Sequential(
nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False)
)
def __init__(self, num_classes=17):
self.inplanes = 128
super(TSnetSE, self).__init__()
self.layer1 = nn.Sequential(
conv3x3(21, self.inplanes),
sn.SwitchNorm2d(self.inplanes),
nn.ReLU(inplace=True),
conv3x3(self.inplanes, self.inplanes),
sn.SwitchNorm2d(self.inplanes),
nn.ReLU(inplace=True),
conv3x3(self.inplanes, self.inplanes, 2),
sn.SwitchNorm2d(self.inplanes),
nn.ReLU(inplace=True),
)
self.se1 = SELayer(self.inplanes)
self.layer2 = nn.Sequential(
conv3x3(self.inplanes, self.inplanes * 2),
sn.SwitchNorm2d(self.inplanes * 2),
nn.ReLU(inplace=True),
conv3x3(self.inplanes * 2, self.inplanes * 2),
sn.SwitchNorm2d(self.inplanes * 2),
nn.ReLU(inplace=True),
conv3x3(self.inplanes * 2, self.inplanes * 2, 2),
sn.SwitchNorm2d(self.inplanes * 2),
nn.ReLU(inplace=True),
)
self.se2 = SELayer(self.inplanes * 2)
self.layer3 = nn.Sequential(
conv3x3(self.inplanes * 2, self.inplanes * 4),
sn.SwitchNorm2d(self.inplanes * 4),
nn.ReLU(inplace=True),
conv3x3(self.inplanes * 4, self.inplanes * 4),
sn.SwitchNorm2d(self.inplanes * 4),
nn.ReLU(inplace=True),
conv3x3(self.inplanes * 4, self.inplanes * 4, 2),
sn.SwitchNorm2d(self.inplanes * 4),
)
self.se3 = SELayer(self.inplanes * 4)
self.fc = nn.Sequential(
nn.Linear(4096 * 2, 1024),
nn.LeakyReLU(inplace=True),
nn.Dropout(p=0.5),
nn.Linear(1024, 512),
nn.LeakyReLU(inplace=True), #relu在drop前面还是后面
nn.Dropout(p=0.8),
nn.Linear(512, 17))
self.LogSoftmax = nn.LogSoftmax()
#后添加
self.initialize()
def __init__(self, in_size, out_size, normalize=True, dropout=0.0):
super(UNetDown, self).__init__()
layers = [nn.Conv2d(in_size, out_size, 4, 2, 1, bias=False)]
if normalize:
layers.append(sn.SwitchNorm2d(out_size))
layers.append(nn.LeakyReLU(0.2))
if dropout:
layers.append(nn.Dropout(dropout))
self.model = nn.Sequential(*layers)
def discriminator_block(in_filters, out_filters, normalization=True):
"""Returns downsampling layers of each discriminator block"""
layers = [
nn.Conv2d(in_filters, out_filters, 4, stride=2, padding=1)
]
if normalization:
layers.append(sn.SwitchNorm2d(out_filters))
layers.append(nn.LeakyReLU(0.2, inplace=True))
return layers
def __init__(self, in_size, out_size, dropout=0.0):
super(UNetUp, self).__init__()
layers = [
nn.ConvTranspose2d(in_size, out_size, 4, 2, 1, bias=False),
sn.SwitchNorm2d(out_size),
nn.ReLU(inplace=True)
]
if dropout:
layers.append(nn.Dropout(dropout))
self.model = nn.Sequential(*layers)