Attention in SPADEGenerator
yf1019 opened this issue · 5 comments
Hi, I notice there is an attention model in the SPADEGenerator, which is defined in the CoCosNet/models/networks/architecture.py line 97. What does this model mean? Can you give a reference? Thanks a lot!
we use attention model to get more globel feature to help synthesis. you can get more information in this paper https://arxiv.org/pdf/1711.07971.pdf
we use attention model to get more globel feature to help synthesis. you can get more information in this paper https://arxiv.org/pdf/1711.07971.pdf
Thanks!
we use attention model to get more globel feature to help synthesis. you can get more information in this paper https://arxiv.org/pdf/1711.07971.pdf
Hi, pan. This work can be used in many fileds. it's great! I just remembered another question.
In the deepfashion dataset, you use the pose landmarks and their joints as input semantics. However, in the code, I noticed this
data/deepfashion.py
It seems you additionly compute one "distance map" for every joint. What is the meaning of this? Is it used to add extra joint information to the model?
the pose image is sparse, only pose landmarks and their joints have nonzero value, so we add distance map to give distance infomantion between zero pixel and joints. This pose representation is better for convolution operator.
the pose image is sparse, only pose landmarks and their joints have nonzero value, so we add distance map to give distance infomantion between zero pixel and joints. This pose representation is better for convolution operator.
I get it. Thanks a lot!