# credits: https://github.com/GeoZcx/A-deeply-supervised-image-fusion-network-for-change-detection-in-remote-sensing-images import torch import torch.nn as nn import torch.nn.functional as F from torchvision.models import vgg16 from opencd.registry import MODELS def get_norm_layer(): # TODO: select appropriate norm layer return nn.BatchNorm2d def get_act_layer(): # TODO: select appropriate activation layer return nn.ReLU def make_norm(*args, **kwargs): norm_layer = get_norm_layer() return norm_layer(*args, **kwargs) def make_act(*args, **kwargs): act_layer = get_act_layer() return act_layer(*args, **kwargs) class BasicConv(nn.Module): def __init__( self, in_ch, out_ch, kernel_size, pad_mode='Zero', bias='auto', norm=False, act=False, **kwargs ): super().__init__() seq = [] if kernel_size >= 2: seq.append(getattr(nn, pad_mode.capitalize()+'Pad2d')(kernel_size//2)) seq.append( nn.Conv2d( in_ch, out_ch, kernel_size, stride=1, padding=0, bias=(False if norm else True) if bias=='auto' else bias, **kwargs ) ) if norm: if norm is True: norm = make_norm(out_ch) seq.append(norm) if act: if act is True: act = make_act() seq.append(act) self.seq = nn.Sequential(*seq) def forward(self, x): return self.seq(x) class Conv1x1(BasicConv): def __init__(self, in_ch, out_ch, pad_mode='Zero', bias='auto', norm=False, act=False, **kwargs): super().__init__(in_ch, out_ch, 1, pad_mode=pad_mode, bias=bias, norm=norm, act=act, **kwargs) class ChannelAttention(nn.Module): def __init__(self, in_ch, ratio=8): super().__init__() self.avg_pool = nn.AdaptiveAvgPool2d(1) self.max_pool = nn.AdaptiveMaxPool2d(1) self.fc1 = Conv1x1(in_ch, in_ch//ratio, bias=False, act=True) self.fc2 = Conv1x1(in_ch//ratio, in_ch, bias=False) def forward(self,x): avg_out = self.fc2(self.fc1(self.avg_pool(x))) max_out = self.fc2(self.fc1(self.max_pool(x))) out = avg_out + max_out return F.sigmoid(out) class SpatialAttention(nn.Module): def __init__(self, kernel_size=7): super().__init__() self.conv = BasicConv(2, 1, kernel_size, bias=False) def forward(self, x): avg_out = torch.mean(x, dim=1, keepdim=True) max_out = torch.max(x, dim=1, keepdim=True)[0] x = torch.cat([avg_out, max_out], dim=1) x = self.conv(x) return F.sigmoid(x) class VGG16FeaturePicker(nn.Module): def __init__(self, indices=(3,8,15,22,29)): super().__init__() features = list(vgg16(pretrained=True).features)[:30] self.features = nn.ModuleList(features).eval() self.indices = set(indices) def forward(self, x): picked_feats = [] for idx, model in enumerate(self.features): x = model(x) if idx in self.indices: picked_feats.append(x) return picked_feats def conv2d_bn(in_ch, out_ch, with_dropout=True): lst = [ nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=1, padding=1), nn.PReLU(), make_norm(out_ch), ] if with_dropout: lst.append(nn.Dropout(p=0.6)) return nn.Sequential(*lst) @MODELS.register_module() class IFN(nn.Module): def __init__(self, use_dropout=False): super().__init__() self.encoder1 = self.encoder2 = VGG16FeaturePicker() self.sa1 = SpatialAttention() self.sa2= SpatialAttention() self.sa3 = SpatialAttention() self.sa4 = SpatialAttention() self.sa5 = SpatialAttention() self.ca1 = ChannelAttention(in_ch=1024) self.bn_ca1 = make_norm(1024) self.o1_conv1 = conv2d_bn(1024, 512, use_dropout) self.o1_conv2 = conv2d_bn(512, 512, use_dropout) self.bn_sa1 = make_norm(512) self.o1_conv3 = Conv1x1(512, 1) self.trans_conv1 = nn.ConvTranspose2d(512, 512, kernel_size=2, stride=2) self.ca2 = ChannelAttention(in_ch=1536) self.bn_ca2 = make_norm(1536) self.o2_conv1 = conv2d_bn(1536, 512, use_dropout) self.o2_conv2 = conv2d_bn(512, 256, use_dropout) self.o2_conv3 = conv2d_bn(256, 256, use_dropout) self.bn_sa2 = make_norm(256) self.o2_conv4 = Conv1x1(256, 1) self.trans_conv2 = nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2) self.ca3 = ChannelAttention(in_ch=768) self.o3_conv1 = conv2d_bn(768, 256, use_dropout) self.o3_conv2 = conv2d_bn(256, 128, use_dropout) self.o3_conv3 = conv2d_bn(128, 128, use_dropout) self.bn_sa3 = make_norm(128) self.o3_conv4 = Conv1x1(128, 1) self.trans_conv3 = nn.ConvTranspose2d(128, 128, kernel_size=2, stride=2) self.ca4 = ChannelAttention(in_ch=384) self.o4_conv1 = conv2d_bn(384, 128, use_dropout) self.o4_conv2 = conv2d_bn(128, 64, use_dropout) self.o4_conv3 = conv2d_bn(64, 64, use_dropout) self.bn_sa4 = make_norm(64) self.o4_conv4 = Conv1x1(64, 1) self.trans_conv4 = nn.ConvTranspose2d(64, 64, kernel_size=2, stride=2) self.ca5 = ChannelAttention(in_ch=192) self.o5_conv1 = conv2d_bn(192, 64, use_dropout) self.o5_conv2 = conv2d_bn(64, 32, use_dropout) self.o5_conv3 = conv2d_bn(32, 16, use_dropout) self.bn_sa5 = make_norm(16) self.o5_conv4 = Conv1x1(16, 1) def forward(self, t1, t2): # Extract bi-temporal features with torch.no_grad(): self.encoder1.eval(), self.encoder2.eval() t1_feats = self.encoder1(t1) t2_feats = self.encoder2(t2) t1_f_l3, t1_f_l8, t1_f_l15, t1_f_l22, t1_f_l29 = t1_feats t2_f_l3, t2_f_l8, t2_f_l15, t2_f_l22, t2_f_l29,= t2_feats # Multi-level decoding x = torch.cat([t1_f_l29, t2_f_l29], dim=1) x = self.o1_conv1(x) x = self.o1_conv2(x) x = self.sa1(x) * x x = self.bn_sa1(x) out1 = self.o1_conv3(x) x = self.trans_conv1(x) x = torch.cat([x, t1_f_l22, t2_f_l22], dim=1) x = self.ca2(x)*x x = self.o2_conv1(x) x = self.o2_conv2(x) x = self.o2_conv3(x) x = self.sa2(x) *x x = self.bn_sa2(x) out2 = self.o2_conv4(x) x = self.trans_conv2(x) x = torch.cat([x, t1_f_l15, t2_f_l15], dim=1) x = self.ca3(x)*x x = self.o3_conv1(x) x = self.o3_conv2(x) x = self.o3_conv3(x) x = self.sa3(x) *x x = self.bn_sa3(x) out3 = self.o3_conv4(x) x = self.trans_conv3(x) x = torch.cat([x, t1_f_l8, t2_f_l8], dim=1) x = self.ca4(x)*x x = self.o4_conv1(x) x = self.o4_conv2(x) x = self.o4_conv3(x) x = self.sa4(x) *x x = self.bn_sa4(x) out4 = self.o4_conv4(x) x = self.trans_conv4(x) x = torch.cat([x, t1_f_l3, t2_f_l3], dim=1) x = self.ca5(x)*x x = self.o5_conv1(x) x = self.o5_conv2(x) x = self.o5_conv3(x) x = self.sa5(x) *x x = self.bn_sa5(x) out5 = self.o5_conv4(x) return (out1, out2, out3, out4, out5)