From ef5aa1ccbb7e1156c6687f40fdf1ae37573c3fb4 Mon Sep 17 00:00:00 2001
From: nagadomi <nagadomi@nurs.or.jp>
Date: Thu, 18 Oct 2018 19:49:10 +0000
Subject: [PATCH] SEBlock

---
 lib/srcnn.lua | 33 ++++++++++++++++++++++++---------
 1 file changed, 24 insertions(+), 9 deletions(-)

diff --git a/lib/srcnn.lua b/lib/srcnn.lua
index 5097b1f..191191c 100644
--- a/lib/srcnn.lua
+++ b/lib/srcnn.lua
@@ -984,7 +984,7 @@ function srcnn.cunet_v4(backend, ch)
    return model
 end
 
-function srcnn.cunet_v5(backend, ch)
+function srcnn.cunet_v6(backend, ch)
    function unet_branch(insert, backend, n_input, n_output, depad)
       local block = nn.Sequential()
       local pooling = SpatialConvolution(backend, n_input, n_input, 2, 2, 2, 2, 0, 0) -- downsampling
@@ -1000,22 +1000,37 @@ function srcnn.cunet_v5(backend, ch)
       model:add(nn.CAddTable())
       return model
    end
-   function unet_conv(n_input, n_middle, n_output)
+   function unet_conv(n_input, n_middle, n_output, se)
 	local model = nn.Sequential()
 	model:add(SpatialConvolution(backend, n_input, n_middle, 3, 3, 1, 1, 0, 0))
 	model:add(nn.LeakyReLU(0.1, true))
 	model:add(SpatialConvolution(backend, n_middle, n_output, 3, 3, 1, 1, 0, 0))
 	model:add(nn.LeakyReLU(0.1, true))
+	if se then
+	   -- Squeeze and Excitation Networks
+	   local con = nn.ConcatTable(2)
+	   local attention = nn.Sequential()
+	   attention:add(nn.SpatialAdaptiveAveragePooling(1, 1)) -- global average pooling
+	   attention:add(SpatialConvolution(backend, n_output, math.floor(n_output / 4), 1, 1, 1, 1, 0, 0))
+	   attention:add(nn.ReLU(true))
+	   attention:add(SpatialConvolution(backend, math.floor(n_output / 4), n_output, 1, 1, 1, 1, 0, 0))
+	   attention:add(nn.Sigmoid(true))
+	   con:add(nn.Identity())                                                                          
+	   con:add(attention)
+	   model:add(con)
+	   model:add(w2nn.ScaleTable())
+	end
 	return model
    end
+   -- Residual U-Net
    function unet(backend, ch, deconv)
-      local block1 = unet_conv(128, 256, 128)
+      local block1 = unet_conv(128, 256, 128, true)
       local block2 = nn.Sequential()
-      block2:add(unet_conv(64, 64, 128))
+      block2:add(unet_conv(64, 64, 128, true))
       block2:add(unet_branch(block1, backend, 128, 128, 4))
-      block2:add(unet_conv(128, 64, 64))
+      block2:add(unet_conv(128, 64, 64, true))
       local model = nn.Sequential()
-      model:add(unet_conv(ch, 32, 64))
+      model:add(unet_conv(ch, 32, 64, false))
       model:add(unet_branch(block2, backend, 64, 64, 16))
       model:add(SpatialConvolution(backend, 64, 64, 3, 3, 1, 1, 0, 0))
       model:add(nn.LeakyReLU(0.1))
@@ -1042,7 +1057,7 @@ function srcnn.cunet_v5(backend, ch)
    model:add(aux_con)
    model:add(w2nn.AuxiliaryLossTable(1)) -- auxiliary loss for single unet output
    
-   model.w2nn_arch_name = "cunet_v5"
+   model.w2nn_arch_name = "cunet_v6"
    model.w2nn_offset = 60
    model.w2nn_scale_factor = 2
    model.w2nn_channels = ch
@@ -1186,13 +1201,13 @@ print(model)
 model:training()
 print(model:forward(torch.Tensor(1, 3, 144, 144):zero():cuda()):size())
 os.exit()
-
-local model = srcnn.cunet_v5("cunn", 3):cuda()
+local model = srcnn.cunet_v6("cunn", 3):cuda()
 print(model)
 model:training()
 print(model:forward(torch.Tensor(1, 3, 144, 144):zero():cuda()))
 os.exit()
 
+
 --]]
 
 return srcnn