def forward(self, x, gts=None):
x_size = x.size() # 800
x0 = self.layer0(x) # 400
x1 = self.layer1(x0) # 400
x2 = self.layer2(x1) # 100
x3 = self.layer3(x2) # 100
x4 = self.layer4(x3) # 100
dec0_up = self.fcn_head(x4)
if self.skip == 'm1':
dec0_fine = self.bot_fine(x1)
dec0_up = Upsample(dec0_up, x1.size()[2:])
else:
dec0_fine = self.bot_fine(x2)
dec0_up = Upsample(dec0_up, x2.size()[2:])
dec0 = [dec0_fine, dec0_up]
dec0 = torch.cat(dec0, 1)
dec1 = self.final(dec0)
main_out = Upsample(dec1, x_size[2:])
if self.training:
return self.criterion(main_out, gts)
return main_out
def forward(self, inp, gts=None):
x_size = inp.size()
x = self.mod1(inp)
m2 = self.mod2(self.pool2(x))
x = self.mod3(self.pool3(m2))
x = self.mod4(x)
x = self.mod5(x)
x = self.mod6(x)
x = self.mod7(x)
x = self.aspp(x)
dec0_up = self.bot_aspp(x)
dec0_fine = self.bot_fine(m2)
dec0_up = Upsample(dec0_up, m2.size()[2:])
dec0 = [dec0_fine, dec0_up]
dec0 = torch.cat(dec0, 1)
dec1 = self.final(dec0)
out = Upsample(dec1, x_size[2:])
if self.training:
return self.criterion(out, gts)
return out
def forward(self, x, gts=None):
x_size = x.size()
x0 = self.layer0(x)
x1 = self.layer1(x0)
x2 = self.layer2(x1)
x3 = self.layer3(x2)
x4 = self.layer4(x3)
x = self.head([x1, x2, x3, x4])
main_out = Upsample(x[0], x_size[2:])
edge_preds = [Upsample(edge_pred, x_size[2:]) for edge_pred in x[1]]
if self.training:
if not self.fpn_dsn:
return self.criterion([main_out, edge_preds], gts)
return self.criterion(x, gts)
return main_out
def forward(self, x, gts=None):
x_size = x.size()[2:] # 800
x0 = self.layer0(x) # 400
x1 = self.layer1(x0) # 400
x2 = self.layer2(x1) # 100
x3 = self.layer3(x2) # 100
x4 = self.layer4(x3) # 100
aux_out = self.aux_layer(x3)
main_out = self.head(x4)
aux_out = Upsample(aux_out, size=x_size)
main_out = Upsample(main_out, size=x_size)
if self.training:
return self.criterion([main_out, aux_out], gts)
return main_out
def forward(self, x, gts=None):
x_size = x.size() # 800
x2, x3, x4 = self.backbone(x)
x = self.head([x2, x3, x4])
main_out = Upsample(x[0], x_size[2:])
if self.training:
if not self.fpn_dsn:
return self.criterion(main_out, gts)
else:
return self.criterion(x, gts)
return main_out
def forward(self, x):
x_size = x.size()
img_features = self.img_pooling(x)
img_features = self.img_conv(img_features)
img_features = Upsample(img_features, x_size[2:])
out = img_features
for f in self.features:
y = f(x)
out = torch.cat((out, y), 1)
return out
def forward(self, x, gts=None):
x_size = x.size() # 800
x0 = self.layer0(x) # 400
x1 = self.layer1(x0) # 400
x2 = self.layer2(x1) # 100
x3 = self.layer3(x2) # 100
x4 = self.layer4(x3) # 100
x = self.head([x1, x2, x3, x4])
main_out = Upsample(x[0], x_size[2:])
if self.training:
if not self.fpn_dsn:
return self.criterion(main_out, gts)
return self.criterion(x, gts)
return main_out
def forward(self, inp, gts=None):
x_size = inp.size()
x = self.mod1(inp)
m2 = self.mod2(self.pool2(x))
fine_size = m2.size()
x = self.mod3(self.pool3(m2))
x = self.mod4(x)
x = self.mod5(x)
x = self.mod6(x)
x = self.mod7(x)
x = self.aspp(x)
aspp = self.bot_aspp(x)
seg_body, seg_edge = self.squeeze_body_edge(aspp)
# may add canny edge
# canny_edge = self.edge_canny(inp, x_size)
# add low-level feature
dec0_fine = self.bot_fine(m2)
seg_edge = self.edge_fusion(
torch.cat([Upsample(seg_edge, fine_size[2:]), dec0_fine], dim=1))
seg_edge_out = self.edge_out(seg_edge)
seg_out = seg_edge + Upsample(seg_body, fine_size[2:])
aspp = Upsample(aspp, fine_size[2:])
seg_out = torch.cat([aspp, seg_out], dim=1)
seg_final = self.final_seg(seg_out)
seg_edge_out = Upsample(seg_edge_out, x_size[2:])
seg_edge_out = self.sigmoid_edge(seg_edge_out)
seg_final_out = Upsample(seg_final, x_size[2:])
seg_body_out = Upsample(self.dsn_seg_body(seg_body), x_size[2:])
# dec0_up = Upsample(dec0_up, m2.size()[2:])
# dec0 = [dec0_fine, dec0_up]
# dec0 = torch.cat(dec0, 1)
# dec1 = self.final(dec0)
# seg_out = Upsample(dec1, x_size[2:])
if self.training:
return self.criterion((seg_final_out, seg_body_out, seg_edge_out),
gts)
return seg_final_out
def forward(self, x, gts=None):
x_size = x.size() # 800
x0 = self.layer0(x) # 400
x1 = self.layer1(x0) # 400
fine_size = x1.size()
x2 = self.layer2(x1) # 100
x3 = self.layer3(x2) # 100
x4 = self.layer4(x3) # 100
xp = self.aspp(x4)
aspp = self.bot_aspp(xp)
seg_body, seg_edge = self.squeeze_body_edge(aspp)
if self.skip == 'm1':
# use default low-level feature
dec0_fine = self.bot_fine(x1)
else:
dec0_fine = self.bot_fine(x2)
seg_edge = self.edge_fusion(
torch.cat([Upsample(seg_edge, fine_size[2:]), dec0_fine], dim=1))
seg_edge_out = self.edge_out(seg_edge)
seg_out = seg_edge + Upsample(seg_body, fine_size[2:])
aspp = Upsample(aspp, fine_size[2:])
seg_out = torch.cat([aspp, seg_out], dim=1)
seg_final = self.final_seg(seg_out)
seg_edge_out = Upsample(seg_edge_out, x_size[2:])
seg_edge_out = self.sigmoid_edge(seg_edge_out)
seg_final_out = Upsample(seg_final, x_size[2:])
seg_body_out = Upsample(self.dsn_seg_body(seg_body), x_size[2:])
if self.training:
return self.criterion((seg_final_out, seg_body_out, seg_edge_out),
gts)
return seg_final_out