kxngluca/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Pull Requests Activity

Allow noise to be added to discriminator inputs

Browse Source
This commit is contained in:
James Betker 2020-05-12 16:25:38 -06:00
parent 9210a62f58
commit 06d18343f7
1 changed files with 17 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
codes/models/SRGAN_model.py
View File

@ -78,6 +78,8 @@ class SRGANModel(BaseModel):
# D_update_ratio and D_init_iters # D_update_ratio and D_init_iters
self.D_update_ratio = train_opt['D_update_ratio'] if train_opt['D_update_ratio'] else 1 self.D_update_ratio = train_opt['D_update_ratio'] if train_opt['D_update_ratio'] else 1
self.D_init_iters = train_opt['D_init_iters'] if train_opt['D_init_iters'] else 0 self.D_init_iters = train_opt['D_init_iters'] if train_opt['D_init_iters'] else 0
self.D_noise_theta = train_opt['D_noise_theta_init'] if train_opt['D_noise_theta_init'] else 0
self.D_noise_final = train_opt['D_noise_final_it'] if train_opt['D_noise_final_it'] else 0
# optimizers # optimizers
# G # G
@ -165,7 +167,14 @@ class SRGANModel(BaseModel):
for p in self.netG.parameters(): for p in self.netG.parameters():
p.requires_grad = False p.requires_grad = False
# Calculate a standard deviation for the gaussian noise to be applied to the discriminator, termed noise-theta.
if step >= self.D_noise_final:
noise_theta = 0
else:
noise_theta = self.D_noise_theta * (self.D_noise_final - step) / self.D_noise_final
self.fake_GenOut = [] self.fake_GenOut = []
var_ref_skips = []
for var_L, var_H, var_ref, pix in zip(self.var_L, self.var_H, self.var_ref, self.pix): for var_L, var_H, var_ref, pix in zip(self.var_L, self.var_H, self.var_ref, self.pix):
fake_GenOut = self.netG(var_L) fake_GenOut = self.netG(var_L)
@ -177,9 +186,11 @@ class SRGANModel(BaseModel):
self.fake_GenOut.append((fake_GenOut[0].detach(), self.fake_GenOut.append((fake_GenOut[0].detach(),
fake_GenOut[1].detach(), fake_GenOut[1].detach(),
fake_GenOut[2].detach())) fake_GenOut[2].detach()))
var_ref = (var_ref,) + self.create_artificial_skips(var_H)
else: else:
gen_img = fake_GenOut gen_img = fake_GenOut
self.fake_GenOut.append(fake_GenOut.detach()) self.fake_GenOut.append(fake_GenOut.detach())
var_ref_skips.append(var_ref)
l_g_total = 0 l_g_total = 0
if step % self.D_update_ratio == 0 and step > self.D_init_iters: if step % self.D_update_ratio == 0 and step > self.D_init_iters:
@ -219,17 +230,13 @@ class SRGANModel(BaseModel):
for p in self.netD.parameters(): for p in self.netD.parameters():
p.requires_grad = True p.requires_grad = True
# Convert var_ref to have the same output format as the generator. This generally means interpolating the noise = torch.randn_like(var_ref[0]) * noise_theta
# HR images to have the same output dimensions as each generator skip connection. noise.to(self.device)
if isinstance(self.fake_GenOut[0], tuple):
var_ref_skips = []
for ref, hi_res in zip(self.var_ref, self.var_H):
var_ref_skips.append((ref,) + self.create_artificial_skips(hi_res))
else:
var_ref_skips = self.var_ref
self.optimizer_D.zero_grad() self.optimizer_D.zero_grad()
for var_L, var_H, var_ref, pix, fake_H in zip(self.var_L, self.var_H, var_ref_skips, self.pix, self.fake_GenOut): for var_L, var_H, var_ref, pix, fake_H in zip(self.var_L, self.var_H, var_ref_skips, self.pix, self.fake_GenOut):
# Apply noise to the inputs to slow discriminator convergence.
var_ref = (var_ref[0] + noise,) + var_ref[1:]
fake_H = (fake_H[0] + noise,) + fake_H[1:]
if self.opt['train']['gan_type'] == 'gan': if self.opt['train']['gan_type'] == 'gan':
# need to forward and backward separately, since batch norm statistics differ # need to forward and backward separately, since batch norm statistics differ
# real # real
@ -297,6 +304,7 @@ class SRGANModel(BaseModel):
self.add_log_entry('l_d_real', l_d_real.item() * self.mega_batch_factor) self.add_log_entry('l_d_real', l_d_real.item() * self.mega_batch_factor)
self.add_log_entry('l_d_fake', l_d_fake.item() * self.mega_batch_factor) self.add_log_entry('l_d_fake', l_d_fake.item() * self.mega_batch_factor)
self.add_log_entry('D_fake', torch.mean(pred_d_fake.detach())) self.add_log_entry('D_fake', torch.mean(pred_d_fake.detach()))
self.add_log_entry('noise_theta', noise_theta)
# Allows the log to serve as an easy-to-use rotating buffer. # Allows the log to serve as an easy-to-use rotating buffer.
def add_log_entry(self, key, value): def add_log_entry(self, key, value):