This commit is contained in:
James Betker 2020-11-30 16:14:21 -07:00
parent 1e0f69e34b
commit a1c8300052
3 changed files with 242 additions and 0 deletions

View File

@ -0,0 +1,86 @@
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
def default_conv(in_channels, out_channels, kernel_size, bias=True):
return nn.Conv2d(
in_channels, out_channels, kernel_size,
padding=(kernel_size//2), bias=bias)
class MeanShift(nn.Conv2d):
def __init__(self, rgb_range, rgb_mean, rgb_std, sign=-1):
super(MeanShift, self).__init__(3, 3, kernel_size=1)
std = torch.Tensor(rgb_std)
self.weight.data = torch.eye(3).view(3, 3, 1, 1)
self.weight.data.div_(std.view(3, 1, 1, 1))
self.bias.data = sign * rgb_range * torch.Tensor(rgb_mean)
self.bias.data.div_(std)
self.requires_grad = False
class BasicBlock(nn.Sequential):
def __init__(
self, in_channels, out_channels, kernel_size, stride=1, bias=False,
bn=True, act=nn.ReLU(True)):
m = [nn.Conv2d(
in_channels, out_channels, kernel_size,
padding=(kernel_size//2), stride=stride, bias=bias)
]
if bn: m.append(nn.BatchNorm2d(out_channels))
if act is not None: m.append(act)
super(BasicBlock, self).__init__(*m)
class ResBlock(nn.Module):
def __init__(
self, conv, n_feats, kernel_size,
bias=True, bn=False, act=nn.ReLU(True), res_scale=1):
super(ResBlock, self).__init__()
m = []
for i in range(2):
m.append(conv(n_feats, n_feats, kernel_size, bias=bias))
if bn: m.append(nn.BatchNorm2d(n_feats))
if i == 0: m.append(act)
self.body = nn.Sequential(*m)
self.res_scale = res_scale
def forward(self, x):
res = self.body(x).mul(self.res_scale)
res += x
return res
class Upsampler(nn.Sequential):
def __init__(self, conv, scale, n_feats, bn=False, act=False, bias=False):
m = []
if (scale & (scale - 1)) == 0: # Is scale = 2^n?
for _ in range(int(math.log(scale, 2))):
m.append(conv(n_feats, 4 * n_feats, 3, bias))
m.append(nn.PixelShuffle(2))
if bn: m.append(nn.BatchNorm2d(n_feats))
if act == 'relu':
m.append(nn.ReLU(True))
elif act == 'prelu':
m.append(nn.PReLU(n_feats))
elif scale == 3:
m.append(conv(n_feats, 9 * n_feats, 3, bias))
m.append(nn.PixelShuffle(3))
if bn: m.append(nn.BatchNorm2d(n_feats))
if act == 'relu':
m.append(nn.ReLU(True))
elif act == 'prelu':
m.append(nn.PReLU(n_feats))
else:
raise NotImplementedError
super(Upsampler, self).__init__(*m)

View File

@ -0,0 +1,152 @@
from models.archs.mdcn import common
import torch
import torch.nn as nn
from utils.util import checkpoint
def make_model(args, parent=False):
return MDCN(args)
class MDCB(nn.Module):
def __init__(self, conv=common.default_conv):
super(MDCB, self).__init__()
n_feats = 128
d_feats = 96
kernel_size_1 = 3
kernel_size_2 = 5
act = nn.ReLU(True)
self.conv_3_1 = conv(n_feats, n_feats, kernel_size_1)
self.conv_3_2 = conv(d_feats, d_feats, kernel_size_1)
self.conv_5_1 = conv(n_feats, n_feats, kernel_size_2)
self.conv_5_2 = conv(d_feats, d_feats, kernel_size_2)
self.confusion_3 = nn.Conv2d(n_feats * 3, d_feats, 1, padding=0, bias=True)
self.confusion_5 = nn.Conv2d(n_feats * 3, d_feats, 1, padding=0, bias=True)
self.confusion_bottle = nn.Conv2d(n_feats * 3 + d_feats * 2, n_feats, 1, padding=0, bias=True)
self.relu = nn.ReLU(inplace=True)
def forward(self, x):
input_1 = x
output_3_1 = self.relu(self.conv_3_1(input_1))
output_5_1 = self.relu(self.conv_5_1(input_1))
input_2 = torch.cat([input_1, output_3_1, output_5_1], 1)
input_2_3 = self.confusion_3(input_2)
input_2_5 = self.confusion_5(input_2)
output_3_2 = self.relu(self.conv_3_2(input_2_3))
output_5_2 = self.relu(self.conv_5_2(input_2_5))
input_3 = torch.cat([input_1, output_3_1, output_5_1, output_3_2, output_5_2], 1)
output = self.confusion_bottle(input_3)
output += x
return output
class CALayer(nn.Module):
def __init__(self, n_feats, reduction=16):
super(CALayer, self).__init__()
# global average pooling: feature --> point
self.avg_pool = nn.AdaptiveAvgPool2d(1)
# feature channel downscale and upscale --> channel weight
self.conv_du = nn.Sequential(
nn.Conv2d(n_feats, n_feats // reduction, 1, padding=0, bias=True),
nn.ReLU(inplace=True),
nn.Conv2d(n_feats // reduction, n_feats, 1, padding=0, bias=True),
nn.Sigmoid()
)
def forward(self, x):
y = self.avg_pool(x)
y = self.conv_du(y)
return x * y
class DB(nn.Module):
def __init__(self, conv=common.default_conv):
super(DB, self).__init__()
n_feats = 128
d_feats = 96
n_blocks = 12
self.fushion_down = nn.Conv2d(n_feats * (n_blocks - 1), d_feats, 1, padding=0, bias=True)
self.channel_attention = CALayer(d_feats)
self.fushion_up = nn.Conv2d(d_feats, n_feats, 1, padding=0, bias=True)
def forward(self, x):
x = self.fushion_down(x)
x = self.channel_attention(x)
x = self.fushion_up(x)
return x
class MDCN(nn.Module):
def __init__(self, args, conv=common.default_conv):
super(MDCN, self).__init__()
n_feats = 128
kernel_size = 3
self.scale_idx = 0
act = nn.ReLU(True)
n_blocks = 12
self.n_blocks = n_blocks
# RGB mean for DIV2K
rgb_mean = (0.4488, 0.4371, 0.4040)
rgb_std = (1.0, 1.0, 1.0)
self.sub_mean = common.MeanShift(args.rgb_range, rgb_mean, rgb_std)
# define head module
modules_head = [conv(args.n_colors, n_feats, kernel_size)]
# define body module
modules_body = nn.ModuleList()
for i in range(n_blocks):
modules_body.append(MDCB())
# define distillation module
modules_dist = nn.ModuleList()
modules_dist.append(DB())
modules_transform = [conv(n_feats, n_feats, kernel_size)]
self.upsample = nn.ModuleList([
common.Upsampler(
conv, s, n_feats, act=True
) for s in args.scale
])
modules_rebult = [conv(n_feats, args.n_colors, kernel_size)]
self.add_mean = common.MeanShift(args.rgb_range, rgb_mean, rgb_std, 1)
self.head = nn.Sequential(*modules_head)
self.body = nn.Sequential(*modules_body)
self.dist = nn.Sequential(*modules_dist)
self.transform = nn.Sequential(*modules_transform)
self.rebult = nn.Sequential(*modules_rebult)
def forward(self, x):
x = self.sub_mean(x)
x = checkpoint(self.head, x)
front = x
MDCB_out = []
for i in range(self.n_blocks):
x = checkpoint(self.body[i], x)
if i != (self.n_blocks - 1):
MDCB_out.append(x)
hierarchical = torch.cat(MDCB_out, 1)
hierarchical = checkpoint(self.dist, hierarchical)
mix = front + hierarchical + x
out = checkpoint(self.transform, mix)
out = self.upsample[self.scale_idx](out)
out = checkpoint(self.rebult, out)
out = self.add_mean(out)
return out
def set_scale(self, scale_idx):
self.scale_idx = scale_idx

View File

@ -173,6 +173,10 @@ def define_G(opt, opt_net, scale=None):
netG = RRDBNet(in_nc=opt_net['in_nc'], out_nc=opt_net['out_nc'], netG = RRDBNet(in_nc=opt_net['in_nc'], out_nc=opt_net['out_nc'],
nf=opt_net['nf'], nb=opt_net['nb'], scale=opt_net['scale'], nf=opt_net['nf'], nb=opt_net['nb'], scale=opt_net['scale'],
initial_conv_stride=opt_net['initial_stride']) initial_conv_stride=opt_net['initial_stride'])
elif which_model == 'mdcn':
from models.archs.mdcn.mdcn import MDCN
args = munchify({'scale': opt_net['scale'], 'n_colors': 3, 'rgb_range': 1.0})
netG = MDCN(args)
else: else:
raise NotImplementedError('Generator model [{:s}] not recognized'.format(which_model)) raise NotImplementedError('Generator model [{:s}] not recognized'.format(which_model))
return netG return netG