weight problem

[litongxin666]

weights = torch.zeros(targets_var.shape)
for i in range(targets_var.shape[0]):
for j in range(targets_var.shape[1]):
if targets_var.data.cpu()[i, j] == -1:
weights[i, j] = weight_neg[j]
elif targets_var.data.cpu()[i, j] == 1:

if targets_var.data.cpu()[i, j] == -1: should be changed to targets_var.data.cpu()[i, j] == 0

[dangweili]

@litongxin666 I re-config the weight in train_deepmar_resnet50.py, line 344-352

[abnercloud]

Hello, @dangweili ,Thanks to the great codebase! I has some confusions as follow:

In train_deepmar_resnet50.py, line 355:
targets_var[targets_var == -1] = 0
What is the meaning of this line?

In evaluate.py, line 45-46:
pt_result[pt_result>=0] = 1
pt_result[pt_result<0] = 0
And, why do we use 0 here?

In demo.py, line 129, 137:
if score[0, idx] >= 0:
Why do we think that greater than 0 means this attribute?

[abnercloud]

criterion = F.binary_cross_entropy_with_logits
After logits, do sigmoid operation. Therefore, 0 is the boundary?

[abnercloud]

What I am confused about is why the label is set to -1 when the data set is loaded.

[abnercloud]

What I am confused about is why the label is set to -1 when the data set is loaded.

Of course, 0 and -1 do not matter.