Welcome to mirror list, hosted at ThFree Co, Russian Federation.

DepthConcat.lua - github.com/torch/nn.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
blob: f64a90eb80aac00cd9ce0cf58517e6c0f735fa0f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
------------------------------------------------------------------------
--[[ DepthConcat ]]--
-- Concatenates the output of Convolutions along the depth dimension
-- (nOutputFrame). This is used to implement the DepthConcat layer
-- of the Going deeper with convolutions paper :
-- http://arxiv.org/pdf/1409.4842v1.pdf
-- The normal Concat Module can't be used since the spatial dimensions
-- of tensors to be concatenated may have different values. To deal with
-- this, we select the largest spatial dimensions and add zero-padding
-- around the smaller dimensions.
------------------------------------------------------------------------
local DepthConcat, _ = torch.class('nn.DepthConcat', 'nn.Concat')

function DepthConcat:windowNarrow(output, currentOutput, offset)
   local outputWindow = output:narrow(self.dimension, offset, currentOutput:size(self.dimension))
   for dim=1,self.outputSize:size(1) do
      local currentSize = currentOutput:size(dim)
      if dim ~= self.dimension and self.outputSize[dim] ~= currentSize then
         -- 5x5 vs 3x3 -> start = [(5-3)/2] + 1 = 2 (1 pad each side)
         -- 9x9 vs 5x5 -> start = [(9-5)/2] + 1 = 3 (2 pad each side)
         -- 9x9 vs 4x4 -> start = [(9-4)/2] + 1 = 3.5 (2 pad, 3 pad)
         local start = math.floor(((self.outputSize[dim] - currentSize) / 2) + 1)
         outputWindow = outputWindow:narrow(dim, start, currentSize)
      end
   end
   return outputWindow
end

function DepthConcat:updateOutput(input)
   self.outputSize = self.outputSize or torch.LongStorage()

   local outs = {}
   for i=1,#self.modules do
      local currentOutput = self:rethrowErrors(self.modules[i], i, 'updateOutput', input)
      outs[i] = currentOutput
      if i == 1 then
         self.outputSize:resize(currentOutput:dim()):copy(currentOutput:size())
      else
         self.outputSize[self.dimension] = self.outputSize[self.dimension] + currentOutput:size(self.dimension)
         for dim=1,self.outputSize:size(1) do
            if dim ~= self.dimension then
               -- take the maximum size (shouldn't change anything for batch dim)
               self.outputSize[dim] = math.max(self.outputSize[dim], currentOutput:size(dim))
            end
         end
      end
   end
   self.output:resize(self.outputSize):zero() --zero for padding

   local offset = 1
   for i,module in ipairs(self.modules) do
      local currentOutput = outs[i]
      local outputWindow = self:windowNarrow(self.output, currentOutput, offset)
      outputWindow:copy(currentOutput)
      offset = offset + currentOutput:size(self.dimension)
   end
   return self.output
end

function DepthConcat:updateGradInput(input, gradOutput)
   self.gradInput:resizeAs(input)

   local offset = 1
   for i,module in ipairs(self.modules) do
      local currentOutput = module.output
      local gradOutputWindow = self:windowNarrow(gradOutput, currentOutput, offset)
      local currentGradInput = self:rethrowErrors(module, i, 'updateGradInput', input, gradOutputWindow)
      if i==1 then
         self.gradInput:copy(currentGradInput)
      else
         self.gradInput:add(currentGradInput)
      end
      offset = offset + currentOutput:size(self.dimension)
   end
   return self.gradInput
end

function DepthConcat:accGradParameters(input, gradOutput, scale)
   scale = scale or 1
   local offset = 1
   for i,module in ipairs(self.modules) do
      local currentOutput = module.output
      local gradOutputWindow = self:windowNarrow(gradOutput, currentOutput, offset)
      self:rethrowErrors(module, i, 'accGradParameters', input, gradOutputWindow, scale)
      offset = offset + currentOutput:size(self.dimension)
   end
end

function DepthConcat:backward(input, gradOutput, scale)
   self.gradInput:resizeAs(input)

   scale = scale or 1
   local offset = 1
   for i,module in ipairs(self.modules) do
      local currentOutput = module.output
      local gradOutputWindow = self:windowNarrow(gradOutput, currentOutput, offset)
      local currentGradInput = self:rethrowErrors(module, i, 'backward', input, gradOutputWindow)
      if i==1 then
         self.gradInput:copy(currentGradInput)
      else
         self.gradInput:add(currentGradInput)
      end
      offset = offset + currentOutput:size(self.dimension)
   end
   return self.gradInput
end

function DepthConcat:accUpdateGradParameters(input, gradOutput, lr)
   local offset = 1
   for i,module in ipairs(self.modules) do
      local currentOutput = module.output
      local gradOutputWindow = self:windowNarrow(gradOutput, currentOutput, offset)
      self:rethrowErrors(module, i, 'accUpdateGradParameters', input, gradOutputWindow, lr)
      offset = offset + currentOutput:size(self.dimension)
   end
end