The code does not match the description of the paper

Thanks for the awesome contribution!

When I read the code, I found that the code did not match the paper.

CSPN/cspn_pytorch/models/torch_resnet_cspn_nyu.py

Line 269 in 24eff12

out = self.relu(self.bn1(self.conv1(x)))

According to the description of your paper, I think we should change

out = self.relu(self.bn1(self.conv1(x)))
out = torch.cat((out, side_input), 1)

to

x = self.relu(self.bn1(self.conv1(x)))
out = torch.cat((x, side_input), 1)

Looking forward to your reply!

Thanks for the awesome contribution!
When I read the code, I found that the code did not match the paper.
  CSPN/cspn_pytorch/models/torch_resnet_cspn_nyu.py


     Line 269
  in
  24eff12





    
      
       out = self.relu(self.bn1(self.conv1(x))) 
According to the description of your paper, I think we should change
out = self.relu(self.bn1(self.conv1(x)))
out = torch.cat((out, side_input), 1)

to
x = self.relu(self.bn1(self.conv1(x)))
out = torch.cat((x, side_input), 1)

Looking forward to your reply!

I wonder is there any difference for the variable

out

in two versions?

The out indeed has no difference in two version. However, x has difference.

According to the description of the paper, x must pass the conv, bn and relu operations before entering the shortcut section.

In my opinion, the two version looks like

and

The out indeed has no difference in two version. However, x has difference.
According to the description of the paper, x must pass the conv, bn and relu operations before entering the shortcut section.
In my opinion, the two version looks like

and

You're right

Thanks for your attention and detailed review.
Yes, this is a mistake, I will update it later.
Thanks