|
|
@@ -1,4 +1,4 @@
|
|
|
-local RandomBinaryCriterion, parent = torch.class('w2nn.RandomBinaryCriterion','nn.Criterion')
|
|
|
+local LBPCriterion, parent = torch.class('w2nn.LBPCriterion','nn.Criterion')
|
|
|
|
|
|
local function create_filters(ch, n, k, layers)
|
|
|
local model = nn.Sequential()
|
|
|
@@ -22,7 +22,7 @@ local function create_filters(ch, n, k, layers)
|
|
|
end
|
|
|
return model
|
|
|
end
|
|
|
-function RandomBinaryCriterion:__init(ch, n, k, layers)
|
|
|
+function LBPCriterion:__init(ch, n, k, layers)
|
|
|
parent.__init(self)
|
|
|
self.layers = layers or 1
|
|
|
self.gamma = 0.1
|
|
|
@@ -38,7 +38,7 @@ function RandomBinaryCriterion:__init(ch, n, k, layers)
|
|
|
self.input = torch.Tensor()
|
|
|
self.target = torch.Tensor()
|
|
|
end
|
|
|
-function RandomBinaryCriterion:updateOutput(input, target)
|
|
|
+function LBPCriterion:updateOutput(input, target)
|
|
|
if input:dim() == 2 then
|
|
|
local k = math.sqrt(input:size(2) / self.ch)
|
|
|
input = input:reshape(input:size(1), self.ch, k, k)
|
|
|
@@ -70,7 +70,7 @@ function RandomBinaryCriterion:updateOutput(input, target)
|
|
|
return self.output
|
|
|
end
|
|
|
|
|
|
-function RandomBinaryCriterion:updateGradInput(input, target)
|
|
|
+function LBPCriterion:updateGradInput(input, target)
|
|
|
local d2 = false
|
|
|
if input:dim() == 2 then
|
|
|
d2 = true
|
nagadomi