search

switchablenorms / DeepFashion_Try_On

Official code for "Towards Photo-Realistic Virtual Try-On by Adaptively Generating↔Preserving Image Content"，CVPR‘20 https://arxiv.org/abs/2003.05863
816 stars 250 forks source link

Second-order constraint constants #87

Closed katiemlewis closed 2 years ago

katiemlewis commented 2 years ago

Hi, could you explain where the 0.02 and 0.08 come from in the STN loss calculation?

rg_loss = torch.max(rg_loss,torch.tensor(0.02).cuda()) cg_loss = torch.max(cg_loss,torch.tensor(0.02).cuda()) rx,ry,cx,cy=torch.tensor(0.08).cuda(),torch.tensor(0.08).cuda()\ ,torch.tensor(0.08).cuda(),torch.tensor(0.08).cuda()

LZQhardworker commented 2 years ago

This is a truncation parameter to allow some flexibility. These are given manually. You can try other parameters