Spaces:
Sleeping
Sleeping
File size: 5,440 Bytes
8b83582 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 | import torch
import torch.nn as nn
# ---------------------------------------------------------------------------
# Original Denoising Autoencoder (96×96 → 96×96)
# ---------------------------------------------------------------------------
class Encoder(nn.Module):
"""Convolutional encoder: 3×96×96 → 256×12×12"""
def __init__(self):
super().__init__()
self.net = nn.Sequential(
# 3×96×96 → 32×96×96
nn.Conv2d(3, 32, kernel_size=3, padding=1),
nn.BatchNorm2d(32),
nn.ReLU(inplace=True),
# 32×96×96 → 64×48×48
nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True),
# 64×48×48 → 128×24×24
nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(128),
nn.ReLU(inplace=True),
# 128×24×24 → 256×12×12
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(256),
nn.ReLU(inplace=True),
)
def forward(self, x):
return self.net(x)
class Decoder(nn.Module):
"""Convolutional decoder: 256×12×12 → 3×96×96"""
def __init__(self):
super().__init__()
self.net = nn.Sequential(
# 256×12×12 → 128×24×24
nn.ConvTranspose2d(256, 128, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(128),
nn.ReLU(inplace=True),
# 128×24×24 → 64×48×48
nn.ConvTranspose2d(128, 64, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True),
# 64×48×48 → 32×96×96
nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(32),
nn.ReLU(inplace=True),
# 32×96×96 → 3×96×96
nn.Conv2d(32, 3, kernel_size=3, padding=1),
nn.Sigmoid(),
)
def forward(self, x):
return self.net(x)
class DenoisingAutoencoder(nn.Module):
"""Full denoising autoencoder: noisy image → clean image."""
def __init__(self):
super().__init__()
self.encoder = Encoder()
self.decoder = Decoder()
def forward(self, x):
z = self.encoder(x)
return self.decoder(z)
def count_parameters(self):
return sum(p.numel() for p in self.parameters() if p.requires_grad)
# ---------------------------------------------------------------------------
# Super-Resolution Autoencoder (noisy 48×48 → clean 96×96)
# ---------------------------------------------------------------------------
class SREncoder(nn.Module):
"""Convolutional encoder: 3×48×48 → 256×6×6"""
def __init__(self):
super().__init__()
self.net = nn.Sequential(
# 3×48×48 → 32×48×48
nn.Conv2d(3, 32, kernel_size=3, padding=1),
nn.BatchNorm2d(32),
nn.ReLU(inplace=True),
# 32×48×48 → 64×24×24
nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True),
# 64×24×24 → 128×12×12
nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(128),
nn.ReLU(inplace=True),
# 128×12×12 → 256×6×6
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.BatchNorm2d(256),
nn.ReLU(inplace=True),
)
def forward(self, x):
return self.net(x)
class SRDecoder(nn.Module):
"""Convolutional decoder: 256×6×6 → 3×96×96 (2× upscale)"""
def __init__(self):
super().__init__()
self.net = nn.Sequential(
# 256×6×6 → 128×12×12
nn.ConvTranspose2d(256, 128, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(128),
nn.ReLU(inplace=True),
# 128×12×12 → 64×24×24
nn.ConvTranspose2d(128, 64, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True),
# 64×24×24 → 32×48×48
nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(32),
nn.ReLU(inplace=True),
# 32×48×48 → 16×96×96 ← extra layer gives the 2× upscale
nn.ConvTranspose2d(32, 16, kernel_size=3, stride=2, padding=1, output_padding=1),
nn.BatchNorm2d(16),
nn.ReLU(inplace=True),
# 16×96×96 → 3×96×96
nn.Conv2d(16, 3, kernel_size=3, padding=1),
nn.Sigmoid(),
)
def forward(self, x):
return self.net(x)
class SuperResAutoencoder(nn.Module):
"""Denoise + 2× upscale: noisy 48×48 → clean 96×96.
Fully convolutional — can accept any input size and will output 2× that size.
"""
def __init__(self):
super().__init__()
self.encoder = SREncoder()
self.decoder = SRDecoder()
def forward(self, x):
return self.decoder(self.encoder(x))
def count_parameters(self):
return sum(p.numel() for p in self.parameters() if p.requires_grad)
|