def forward(self, x):
# Encoder
x, pad = pad_image_tensor(x, 32)
x = self.firstconv(x)
x = self.firstbn(x)
x = self.firstrelu(x)
x = self.firstmaxpool(x)
e1 = self.encoder1(x)
e2 = self.encoder2(e1)
e3 = self.encoder3(e2)
e4 = self.encoder4(e3)
# Decoder with Skip Connections
d4 = self.decoder4(e4) + e3
d3 = self.decoder3(d4) + e2
d2 = self.decoder2(d3) + e1
d1 = self.decoder1(d2)
d1 = self.finaldropout(d1)
# Final Classification
f1 = self.finaldeconv1(d1)
f2 = self.finalrelu1(f1)
f3 = self.finalconv2(f2)
f4 = self.finalrelu2(f3)
f5 = self.finalconv3(f4)
f5 = unpad_image_tensor(f5, pad)
return f5
def test_pad_unpad_nonsymmetric(shape, padding):
x = torch.randn(shape)
x_padded, pad_params = pad_image_tensor(x, pad_size=padding)
assert x_padded.size(2) % padding[0] == 0
assert x_padded.size(3) % padding[1] == 0
y = unpad_image_tensor(x_padded, pad_params)
assert (x == y).all()
def forward(self, x):
x, pad = pad_image_tensor(x, 32)
enc_features = self.encoder(x)
dec_features = self.decoder(enc_features)
features = self.fpn_fuse(dec_features)
features = self.dropout(features)
features = self.final_decoder(features)
logits = self.logits(features)
logits = F.interpolate(logits, size=(x.size(2), x.size(3)), mode='bilinear', align_corners=True)
logits = unpad_image_tensor(logits, pad)
return logits
def forward(self, x):
x, pad = pad_image_tensor(x, 32)
enc_features = self.encoder(x)
# Decode mask
mask, dsv = self.decoder(enc_features)
if self.full_size_mask:
mask = F.interpolate(mask,
size=x.size()[2:],
mode="bilinear",
align_corners=False)
mask = unpad_image_tensor(mask, pad)
output = {OUTPUT_MASK_KEY: mask, OUTPUT_MASK_32_KEY: dsv}
return output