def __init__(self, in_nc=3, out_nc=3, nf=64, nb=16, upscale=4): super(MSRResNet, self).__init__() self.upscale = upscale self.conv_first = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True) basic_block = functools.partial(mutil.ResidualBlock_noBN, nf=nf) self.recon_trunk = mutil.make_layer(basic_block, nb) # upsampling if self.upscale == 2: self.upconv1 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(2) elif self.upscale == 3: self.upconv1 = nn.Conv2d(nf, nf * 9, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(3) elif self.upscale == 4: self.upconv1 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.upconv2 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(2) self.HRconv = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.conv_last = nn.Conv2d(nf, out_nc, 3, 1, 1, bias=True) # activation function self.lrelu = nn.LeakyReLU(negative_slope=0.1, inplace=True) # initialization mutil.initialize_weights( [self.conv_first, self.upconv1, self.HRconv, self.conv_last], 0.1) if self.upscale == 4: mutil.initialize_weights(self.upconv2, 0.1)
def __init__(self, nf=64, gc=32, bias=True, use_snorm=False): super(ResidualDenseBlock_5C, self).__init__() # gc: growth channel, i.e. intermediate channels if use_snorm: self.conv1 = nn.utils.spectral_norm( nn.Conv2d(nf, gc, 3, 1, 1, bias=bias)) self.conv2 = nn.utils.spectral_norm( nn.Conv2d(nf + gc, gc, 3, 1, 1, bias=bias)) self.conv3 = nn.utils.spectral_norm( nn.Conv2d(nf + 2 * gc, gc, 3, 1, 1, bias=bias)) self.conv4 = nn.utils.spectral_norm( nn.Conv2d(nf + 3 * gc, gc, 3, 1, 1, bias=bias)) self.conv5 = nn.utils.spectral_norm( nn.Conv2d(nf + 4 * gc, nf, 3, 1, 1, bias=bias)) else: self.conv1 = nn.Conv2d(nf, gc, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(nf + gc, gc, 3, 1, 1, bias=bias) self.conv3 = nn.Conv2d(nf + 2 * gc, gc, 3, 1, 1, bias=bias) self.conv4 = nn.Conv2d(nf + 3 * gc, gc, 3, 1, 1, bias=bias) self.conv5 = nn.Conv2d(nf + 4 * gc, nf, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) # initialization mutil.initialize_weights( [self.conv1, self.conv2, self.conv3, self.conv4, self.conv5], 0.1)
def __init__( self, channel_in: int, channel_out: int, init: str = "xavier", gc: int = 32, bias=True, ): # https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html super(DenseBlock, self).__init__() self.conv1 = nn.Conv2d(channel_in, gc, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(channel_in + gc, gc, 3, 1, 1, bias=bias) self.conv3 = nn.Conv2d(channel_in + 2 * gc, gc, 3, 1, 1, bias=bias) self.conv4 = nn.Conv2d(channel_in + 3 * gc, gc, 3, 1, 1, bias=bias) self.conv5 = nn.Conv2d(channel_in + 4 * gc, channel_out, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) if init == "xavier": mutil.initialize_weights_xavier( [self.conv1, self.conv2, self.conv3, self.conv4], 0.1) else: mutil.initialize_weights( [self.conv1, self.conv2, self.conv3, self.conv4], 0.1) mutil.initialize_weights(self.conv5, 0)
def __init__(self, channel_in, channel_out, init='xavier', gc=32, bias=True): super(DenseBlock, self).__init__() self.conv1 = nn.Conv2d(channel_in, gc, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(channel_in + gc, gc, 3, 1, 1, bias=bias) self.conv3 = nn.Conv2d(channel_in + 2 * gc, gc, 3, 1, 1, bias=bias) self.conv4 = nn.Conv2d(channel_in + 3 * gc, gc, 3, 1, 1, bias=bias) self.conv5 = nn.Conv2d(channel_in + 4 * gc, channel_out, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) if init == 'xavier': mutil.initialize_weights_xavier( [self.conv1, self.conv2, self.conv3, self.conv4], 0.1) else: mutil.initialize_weights( [self.conv1, self.conv2, self.conv3, self.conv4], 0.1) mutil.initialize_weights(self.conv5, 0)
def __init__(self, filters=64, bias=True): super(ResidualDenseBlock_2C, self).__init__() # gc: growth channel, i.e. intermediate channels self.conv1 = nn.Conv2d(filters, filters, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(filters, filters, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) # initialization mutil.initialize_weights([self.conv1, self.conv2], 0.1)
def __init__(self, nf=64, gc=32, bias=True): super(ResidualDenseBlock_5C, self).__init__() self.conv1 = nn.Conv2d(nf, gc, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(nf + gc, gc, 3, 1, 1, bias=bias) self.conv3 = nn.Conv2d(nf + 2 * gc, gc, 3, 1, 1, bias=bias) self.conv4 = nn.Conv2d(nf + 3 * gc, gc, 3, 1, 1, bias=bias) self.conv5 = nn.Conv2d(nf + 4 * gc, nf, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) mutil.initialize_weights( [self.conv1, self.conv2, self.conv3, self.conv4, self.conv5], 0.1)
def __init__(self, nf=64, gc=32, bias=True): super(TimeResidualDenseBlock5C, self).__init__() # gc: growth channel, i.e. intermediate channels self.conv1 = ConcatConv2d(nf, gc, 3, 1, 1, bias=bias) self.conv2 = ConcatConv2d(nf + gc, gc, 3, 1, 1, bias=bias) self.conv3 = ConcatConv2d(nf + 2 * gc, gc, 3, 1, 1, bias=bias) self.conv4 = ConcatConv2d(nf + 3 * gc, gc, 3, 1, 1, bias=bias) self.conv5 = ConcatConv2d(nf + 4 * gc, nf, 3, 1, 1, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) # initialization mutil.initialize_weights([self.conv1, self.conv2, self.conv3, self.conv4, self.conv5], 0.1)
def __init__(self, in_nc, out_nc, nf, nb, gc=32, differential=None, time_dependent=False, adjoint=False, sb=5): super(RRDBNet, self).__init__() self.conv_first = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True) self.trunk_conv = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) if differential == "checkpointed": self.conv_trunk = SRTrunk(nf, nb, make_odeblock(5, 'RK4')) mutil.initialize_weights(self.conv_trunk.odefunc.convs) elif differential == "standard": self.conv_trunk = ODEBlock(ODEfunc(nf, nb=nb, normalization=False, time_dependent=time_dependent), adjoint=adjoint) mutil.initialize_weights(self.conv_trunk.odefunc.convs) elif differential == "sequential": self.conv_trunk = nn.Sequential(*[ODEBlock(ODEfunc(nf, nb=sb, normalization=False, time_dependent=time_dependent), adjoint=adjoint) for _ in range(nb)]) for block in self.conv_trunk: mutil.initialize_weights(block.odefunc.convs) elif differential == "augmented": augment_dim = nf//4 method = 'dopri5' if method == 'euler': warnings.warn("euler mode") self.conv_trunk = AugBlock(AugFunc(nf=nf, nb=nb, augment_dim=augment_dim, time_dependent=time_dependent), adjoint=adjoint, is_conv=True, method=method) self.trunk_conv = nn.Conv2d(nf+augment_dim, nf, 3, 1, 1, bias=True) mutil.initialize_weights(self.conv_trunk.odefunc.convs) elif differential is None or differential == "nodiff": RRDB_block_f = functools.partial(RRDB, nf=nf, gc=gc) self.conv_trunk = mutil.make_layer(RRDB_block_f, nb) else: raise NotImplementedError("unrecognized differential system passed") #### upsampling self.upconv1 = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.upconv2 = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.HRconv = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.conv_last = nn.Conv2d(nf, out_nc, 3, 1, 1, bias=True) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)
def __init__(self, in_nc=3, out_nc=3, nf=64, nb=16): super(MSRResNet, self).__init__() self.conv_first = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True) basic_block = functools.partial(mutil.ResidualBlock_noBN, nf=nf) self.recon_trunk = mutil.make_layer(basic_block, nb) self.HRconv = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.conv_last = nn.Conv2d(nf, out_nc, 3, 1, 1, bias=True) # activation function self.lrelu = nn.LeakyReLU(negative_slope=0.1, inplace=True) # initialization mutil.initialize_weights( [self.conv_first, self.HRconv, self.conv_last], 0.1)
def __init__(self, nf=64, gc=32, bias=True, use_snorm=False): super(Multi_extfea, self).__init__() # gc: growth channel, i.e. intermediate channels if use_snorm: self.conv0 = nn.utils.spectral_norm( nn.Conv2d(nf, gc, 1, 1, 1, bias=bias)) self.conv1 = nn.utils.spectral_norm( nn.Conv2d(nf, gc, 3, 1, 1, bias=bias)) self.conv2 = nn.utils.spectral_norm( nn.Conv2d(nf, gc, 5, 1, 2, bias=bias)) self.conv3 = nn.utils.spectral_norm( nn.Conv2d(nf, gc, 7, 1, 3, bias=bias)) else: self.conv0 = nn.Conv2d(nf, gc, 1, 1, 1, bias=bias) self.conv1 = nn.Conv2d(nf, gc, 3, 1, 1, bias=bias) self.conv2 = nn.Conv2d(nf, gc, 5, 1, 2, bias=bias) self.conv3 = nn.Conv2d(nf, gc, 7, 1, 3, bias=bias) self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) self.conv_out = nn.Conv2d(gc * 4, gc, 3, 1, 1, bias=bias) # initialization mutil.initialize_weights( [self.conv0, self.conv1, self.conv2, self.conv3, self.conv_out], 0.1)
def __init__(self, in_nc=3, out_nc=3, nf=64, nb=16, upscale=4, differential=False): super(MSRResNet, self).__init__() self.upscale = upscale self.conv_first = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True) if differential == "checkpointed": self.conv_trunk = SRTrunk(nf, nb, make_odeblock(2, 'RK4')) mutil.initialize_weights(self.conv_trunk.odefunc.convs) elif differential == "standard": self.conv_trunk = ODEBlock(ODEfunc(nf, nb=nb, normalization=False)) mutil.initialize_weights(self.conv_trunk.odefunc.convs) elif differential is None: basic_block = functools.partial(mutil.ResidualBlock_noBN, nf=nf) self.conv_trunk = mutil.make_layer(basic_block, nb) else: raise NotImplementedError( "unrecognized differential system passed") # upsampling if self.upscale == 2: self.upconv1 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(2) elif self.upscale == 3: self.upconv1 = nn.Conv2d(nf, nf * 9, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(3) elif self.upscale == 4: self.upconv1 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.upconv2 = nn.Conv2d(nf, nf * 4, 3, 1, 1, bias=True) self.pixel_shuffle = nn.PixelShuffle(2) self.HRconv = nn.Conv2d(nf, nf, 3, 1, 1, bias=True) self.conv_last = nn.Conv2d(nf, out_nc, 3, 1, 1, bias=True) # activation function self.lrelu = nn.LeakyReLU(negative_slope=0.1, inplace=False) # initialization mutil.initialize_weights( [self.conv_first, self.upconv1, self.HRconv, self.conv_last], 0.1) if self.upscale == 4: mutil.initialize_weights(self.upconv2, 0.1)