123456/roma_unsb
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use kun's ctn method

Browse Source
This commit is contained in:
bishe 2025-02-23 21:15:41 +08:00
parent 6705075876
commit 0639032b6c
3 changed files with 5 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
checkpoints/ROMA_UNSB_001/loss_log.txt
View File

@ -33,3 +33,4 @@
================ Training Loss (Sun Feb 23 18:59:52 2025) ================
================ Training Loss (Sun Feb 23 19:03:05 2025) ================
================ Training Loss (Sun Feb 23 19:03:57 2025) ================
================ Training Loss (Sun Feb 23 21:11:47 2025) ================

BIN
models/__pycache__/roma_unsb_model.cpython-39.pyc
View File

Binary file not shown.

17
models/roma_unsb_model.py
View File

@ -527,26 +527,17 @@ class RomaUnsbModel(BaseModel):
setattr(self, "fake_"+str(t+1), Xt_1)
if self.opt.phase == 'train':
print(f'real_B0.shape = {real_B0.shape} fake_B0.shape = {self.fake_B0.shape}')
print(f"self.real_B0.requires_grad: {real_B0.requires_grad}")
# 真实图像的梯度
real_gradient = torch.autograd.grad(real_B0.sum(), real_B0, create_graph=True)[0]
# 生成图像的梯度
fake_gradient = torch.autograd.grad(self.fake_B0.sum(), self.fake_B0, create_graph=True)[0]
# 梯度图
self.weight_real, self.weight_fake = self.cao.generate_weight_map(fake_gradient)
self.weight_fake = self.cao.generate_weight_map(fake_gradient)
# 生成图像的CTN光流图
self.f_content = self.ctn(self.weight_fake)
# 把前面生成后的图片再加上noisy_map
self.fake_B_2 = self.fake_B + self.noisy_map
# 变换后的图片
wapped_fake_B = warp(self.fake_B, self.f_content)
self.warped_real_A_noisy2 = warp(self.real_A_noisy, self.f_content)
self.warped_fake_B0 = warp(self.fake_B0,self.f_content)
# 经过第二次生成器
self.fake_B_2 = self.netG(wapped_fake_B, self.time, z_in)
self.warped_fake_B0_2 = self.netG(self.warped_real_A_noisy2, self.time, z_in)
def compute_D_loss(self):
"""计算判别器的 GAN 损失"""