search

HsinYingLee / DRIT

Learning diverse image-to-image translation from unpaired data
842 stars 154 forks source link

code about self-reconstuction loss ? #47

Open jiaying96 opened 4 years ago

jiaying96 commented 4 years ago

Are these code about self-reconstuction loss ?

loss_G_L1_A = self.criterionL1(self.fake_A_recon, self.real_A_encoded)
 loss_G_L1_B = self.criterionL1(self.fake_B_recon, self.real_B_encoded)

AND

Are these code about cross cycle consistency loss ?

loss_G_L1_AA = self.criterionL1(self.fake_AA_encoded, self.real_A_encoded)
    loss_G_L1_BB = self.criterionL1(self.fake_BB_encoded, self.real_B_encoded)
hytseng0509 commented 4 years ago

Correct.