def __init__(self, channels): super(EncodeNet, self).__init__() self.conv_channels_up = nn.Conv2d(1, channels, 1) convList = [] for i in range(4): convList.append(pixelShuffleFunction.PixelUnshuffle(2)) if(i>=2): convList.append(nn.Conv2d(channels * 4, channels, 1)) elif(i==0): convList.append(nn.Conv2d(channels * 4, channels, 5, padding=2)) elif(i==1): convList.append(nn.Conv2d(channels * 4, channels, 3, padding=1)) ''' i kernel padding 0 5 2 1 3 1 2 1 0 3 1 0 ''' convList.append(nn.LeakyReLU(inplace=True)) convList.append(nn.BatchNorm2d(channels)) convList.append(baseNet.ResNet(transpose=False, channels=channels, kernel_size=3, padding=1)) convList.append(nn.Conv2d(channels, 16, 1)) convList.append(nn.LeakyReLU(inplace=True)) self.convList = nn.Sequential(*convList)
def __init__(self, channels): super(EncodeNet, self).__init__() self.conv_channels_up = nn.Conv2d(1, channels, 1) convList = [] for i in range(4): convList.append(pixelShuffleFunction.PixelUnshuffle(2)) convList.append(nn.Conv2d(channels*4, channels, 1)) convList.append(nn.LeakyReLU()) convList.append(nn.Conv2d(channels, 16, 1)) convList.append(nn.LeakyReLU()) self.convList = nn.Sequential(*convList)
def __init__(self, channels, downSample=True): super(SampleResNet, self).__init__() self.downSample = downSample convList = [] if(self.downSample==True): convList.append(pixelShuffleFunction.PixelUnshuffle(2)) convList.append(nn.Conv2d(channels * 4, channels, 1)) convList.append(nn.LeakyReLU()) else: convList.append(nn.Conv2d(channels, channels * 4, 1)) convList.append(nn.LeakyReLU()) convList.append(nn.PixelShuffle(2)) self.convList = nn.Sequential(*convList)
def __init__(self, channels): super(EncodeNet, self).__init__() self.conv_channels_up = nn.Conv2d(1, channels, 1) convList = [] for i in range(4): convList.append(pixelShuffleFunction.PixelUnshuffle(2)) if(i<=1): convList.append(nn.Conv2d(channels*4, channels, 3, padding=1)) else: convList.append(nn.Conv2d(channels * 4, channels, 1)) convList.append(nn.LeakyReLU(inplace=True)) convList.append(nn.BatchNorm2d(channels)) convList.append(nn.Conv2d(channels, 16, 1)) convList.append(nn.LeakyReLU(inplace=True)) self.convList = nn.Sequential(*convList)
def __init__(self, channels): super(EncodeNet, self).__init__() self.conv_channels_up = nn.Conv2d(1, channels, 1) convList = [] for i in range(4): convList.append(pixelShuffleFunction.PixelUnshuffle(2)) convList.append(nn.Conv2d(channels * 4, channels, 1)) convList.append(nn.LeakyReLU()) convList.append( baseNet.ResNet(transpose=False, channels=channels, kernel_size=3, padding=1)) convList.append(nn.Conv2d(channels, 16, 1)) convList.append(nn.LeakyReLU()) self.convList = nn.Sequential(*convList)