Merge pull request #138 from DerThorsten/master

fixed unet examples

examples/res_unet.py

@@ -7,9 +7,7 @@
 
 """
 import torch
-import inferno.extensions.layers.unet as unet
-
-from inferno.extensions.layers import ConvELU2D, ConvSigmoid2D
+from inferno.extensions.layers import (ResBlockUNet, UNetBase, ConvELU2D, ConvSigmoid2D)
 
 
 
@@ -19,7 +17,7 @@
 x = torch.autograd.Variable(x)
 
 # a unet with resiudal blocks
-model = unet.ResBlockUNet(in_channels=10, out_channels=20, dim=2)
+model = ResBlockUNet(in_channels=10, out_channels=20, dim=2)
 
 # pass x trough unet 
 out = model(x)
@@ -43,14 +41,14 @@
 x = torch.autograd.Variable(x)
 
 # a unet with resiudal blocks
-model_a = unet.ResBlockUNet(in_channels=5, out_channels=12, dim=3, 
+model_a = ResBlockUNet(in_channels=5, out_channels=12, dim=3, 
                             unet_kwargs=dict(depth=3))
 
 # if the last layer in the second unet
 # shall be non-activated  we set 
 # activated to False, this will only affect the
 # very last convolution of the net
-model_b = unet.ResBlockUNet(in_channels=12, out_channels=2,  dim=3,
+model_b = ResBlockUNet(in_channels=12, out_channels=2,  dim=3,
                             activated=False,
                             unet_kwargs=dict(depth=3))
 
@@ -72,7 +70,7 @@
 a custom UNet by deriving from UNetBase
 """
 
-class MySimple2DUnet(unet.UNetBase):
+class MySimple2DUnet(UNetBase):
     def __init__(self, **kwargs):
         super(MySimple2DUnet, self).__init__(dim=2, **kwargs)
 
@@ -85,13 +83,13 @@ def conv_op_factory(self, in_channels, out_channels, part, index):
                 ConvELU2D(in_channels=in_channels,  out_channels=out_channels, kernel_size=3),
                 ConvELU2D(in_channels=out_channels, out_channels=out_channels, kernel_size=3),
                 ConvSigmoid2D(in_channels=out_channels, out_channels=out_channels, kernel_size=3)
-            )
+            ), False
         else:
             return torch.nn.Sequential(
                 ConvELU2D(in_channels=in_channels,  out_channels=out_channels, kernel_size=3),
                 ConvELU2D(in_channels=out_channels, out_channels=out_channels, kernel_size=3),
                 ConvELU2D(in_channels=out_channels, out_channels=out_channels, kernel_size=3)
-            )
+            ), False
 
 
 

examples/train_side_loss_unet.py

@@ -16,7 +16,7 @@
 
 from inferno.extensions.layers.convolutional import  Conv2D
 from inferno.extensions.layers.building_blocks import ResBlock
-from inferno.extensions.layers.unet import ResBlockUNet
+from inferno.extensions.layers import ResBlockUNet
 from inferno.utils.torch_utils import unwrap
 from inferno.utils.python_utils import ensure_dir
 import pylab

tests/extensions/layers/unet.py

@@ -3,7 +3,7 @@
 from inferno.extensions.layers import UNetBase
 from inferno.extensions.layers import ResBlockUNet
 from inferno.extensions.layers.convolutional import ConvELU2D
-
+-