It seems that the output stack in Line 92 haven't been used in the rest of the network, is that right?
Also, it looks like every output (stack of Line 108) of DBs in upsample path were unused, since this line:
stack = TransitionUp(skip_connection_list[i], block_to_upsample, n_filters_keep)
TU overwrite stack every time after a DB output that stack .
If the assumption above is correct, the m of first upsampling should be 656 + 12x16 = 848
Hi, I've got a question here:
In your code here: https://github.com/SimJeg/FC-DenseNet/blob/master/FC-DenseNet.py#L98-L108
It seems that the output
stack
in Line 92 haven't been used in the rest of the network, is that right?Also, it looks like every output (
stack
of Line 108) of DBs in upsample path were unused, since this line:stack = TransitionUp(skip_connection_list[i], block_to_upsample, n_filters_keep)
TU overwritestack
every time after a DB output thatstack
.If the assumption above is correct, the
m
of first upsampling should be 656 + 12x16 = 848Do I miss something?