SpatialAveragePooling divides by kW*kH

4 files changed, 19 insertions, 6 deletions

diff --git a/SpatialAveragePooling.lua b/SpatialAveragePooling.lua
index 13b6b45..90b79aa 100644
--- a/SpatialAveragePooling.lua
+++ b/SpatialAveragePooling.lua
@@ -7,14 +7,26 @@ function SpatialAveragePooling:__init(kW, kH, dW, dH)
    self.kH = kH
    self.dW = dW or 1
    self.dH = dH or 1
+   self.divide = true
 end
 
 function SpatialAveragePooling:updateOutput(input)
-   return input.nn.SpatialAveragePooling_updateOutput(self, input)
+   input.nn.SpatialAveragePooling_updateOutput(self, input)
+   -- for backward compatibility with saved models
+   -- which are not supposed to have "divide" field
+   if not self.divide then
+     self.output:mul(self.kW*self.kH)
+   end
+   return self.output
 end
 
 function SpatialAveragePooling:updateGradInput(input, gradOutput)
    if self.gradInput then
-      return input.nn.SpatialAveragePooling_updateGradInput(self, input, gradOutput)
+      input.nn.SpatialAveragePooling_updateGradInput(self, input, gradOutput)
+      -- for backward compatibility
+      if not self.divide then
+	self.gradInput:mul(self.kW*self.kH)
+      end
+      return self.gradInput
    end
 end
diff --git a/SpatialLPPooling.lua b/SpatialLPPooling.lua
index 4e6fe74..fc56296 100644
--- a/SpatialLPPooling.lua
+++ b/SpatialLPPooling.lua
@@ -17,6 +17,7 @@ function SpatialLPPooling:__init(nInputPlane, pnorm, kW, kH, dW, dH)
       self:add(nn.Power(pnorm))
    end
    self:add(nn.SpatialAveragePooling(kW, kH, dW, dH))
+   self:add(nn.MulConstant(kW*kH))
    if pnorm == 2 then
       self:add(nn.Sqrt())
    else
diff --git a/generic/SpatialAveragePooling.c b/generic/SpatialAveragePooling.c
index 2052d05..681938e 100644
--- a/generic/SpatialAveragePooling.c
+++ b/generic/SpatialAveragePooling.c
@@ -83,7 +83,7 @@ static int nn_(SpatialAveragePooling_updateOutput)(lua_State *L)
             ptr_input += inputWidth; /* next input line */
           }
           /* Update output */
-          *ptr_output++ += sum;
+          *ptr_output++ += sum/(kW*kH);
         }
       }
     }
@@ -163,7 +163,7 @@ static int nn_(SpatialAveragePooling_updateGradInput)(lua_State *L)
           for(ky = 0; ky < kH; ky++)
           {
             for(kx = 0; kx < kW; kx++)
-              ptr_gradInput[kx] += z;
+              ptr_gradInput[kx] += z/(kW*kH);
             ptr_gradInput += inputWidth;
           }
         }
diff --git a/test.lua b/test.lua
index f1576fc..8f5fbaa 100644
--- a/test.lua
+++ b/test.lua
@@ -1706,7 +1706,7 @@ function nntest.SpatialAveragePooling()
    mytester:asserteq(berr, 0, torch.typename(module) .. ' - i/o backward err ')
    
    local sap = nn.SpatialSubSampling(from, ki, kj, si, sj)
-   sap.weight:fill(1.0)
+   sap.weight:fill(1.0/(ki*kj))
    sap.bias:fill(0.0)
    
    local output = module:forward(input)
@@ -1737,7 +1737,7 @@ function nntest.SpatialAveragePooling()
    mytester:asserteq(berr, 0, torch.typename(module) .. ' - i/o backward err (Batch) ')
    
    local sap = nn.SpatialSubSampling(from, ki, kj, si, sj)
-   sap.weight:fill(1.0)
+   sap.weight:fill(1.0/(ki*kj))
    sap.bias:fill(0.0)
    
    local output = module:forward(input)