Regarding segnet

[sriharsha0806]

Hi,
Regarding segnet upsasmpling segnetup3(), segnetup2(). shouldn't the code be

class segnetUp2(nn.Module):
def init(self, in_size, out_size):
super(segnetUp2, self).init()
self.unpool = nn.MaxUnpool2d(2, 2)
self.conv1 = conv2DBatchNormRelu(in_size, in_size, 3, 1, 1)
self.conv2 = conv2DBatchNormRelu(in_size, out_size, 3, 1, 1)

def forward(self, inputs, indices, output_shape):
    outputs = self.unpool(input=inputs, indices=indices, output_size=output_shape)
    outputs = self.conv1(outputs)
    outputs = self.conv2(outputs)
    return outputs

class segnetUp3(nn.Module):
def init(self, in_size, out_size):
super(segnetUp3, self).init()
self.unpool = nn.MaxUnpool2d(2, 2)
self.conv1 = conv2DBatchNormRelu(in_size, in_size, 3, 1, 1)
self.conv2 = conv2DBatchNormRelu(int_size, in_size, 3, 1, 1)
self.conv3 = conv2DBatchNormRelu(in_size, out_size, 3, 1, 1)

Because the Decoder Architecture according to the program is
Deconv3-512X3->Deconv3-256X3--->Deconv3-128X3--->Deconv3-64X2--->Deconv3-21x2

But the actual architecture is
Deconv3-512X3->Deconv3-512X3--->Deconv3-256X3--->Deconv3-128X2--->Deconv3-64x2

[meetps]

Yup, thanks!

1b7d852..792d4c1 fixes it.