-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
135 lines (103 loc) · 4.42 KB
/
utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
import os
from os.path import join as ospj
import json
import glob
from shutil import copyfile
from tqdm import tqdm
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
import torchvision.utils as vutils
def save_json(json_file, filename):
with open(filename, 'w') as f:
json.dump(json_file, f, indent=4, sort_keys=False)
def print_network(network, name):
num_params = 0
for p in network.parameters():
num_params += p.numel()
# print(network)
print("Number of parameters of %s: %i" % (name, num_params))
def he_init(module):
if isinstance(module, nn.Conv2d):
nn.init.kaiming_normal_(module.weight, mode='fan_in', nonlinearity='relu')
if module.bias is not None:
nn.init.constant_(module.bias, 0)
if isinstance(module, nn.Linear):
nn.init.kaiming_normal_(module.weight, mode='fan_in', nonlinearity='relu')
if module.bias is not None:
nn.init.constant_(module.bias, 0)
def denorm(x):
out = (x + 1) / 2
return out.clamp_(0, 1)
def save_image(x, ncol, filename):
x = denorm(x)
vutils.save_image(x.cpu(), filename, nrow=ncol, padding=0)
@torch.no_grad()
def translate_and_reconstruct(nets, args, x_src, y_src, x_ref, y_ref, filename):
N, C, H, W = x_src.size()
s_ref = nets.style_encoder(x_ref, y_ref)
masks = nets.fan.get_heatmap(x_src) if args.w_hpf > 0 else None
x_fake = nets.generator(x_src, s_ref, masks=masks)
s_src = nets.style_encoder(x_src, y_src)
masks = nets.fan.get_heatmap(x_fake) if args.w_hpf > 0 else None
x_rec = nets.generator(x_fake, s_src, masks=masks)
x_concat = [x_src, x_ref, x_fake, x_rec]
x_concat = torch.cat(x_concat, dim=0)
save_image(x_concat, N, filename)
del x_concat
@torch.no_grad()
def translate_using_latent(nets, args, x_src, y_trg_list, z_trg_list, psi, filename):
N, C, H, W = x_src.size()
latent_dim = z_trg_list[0].size(1)
x_concat = [x_src]
masks = nets.fan.get_heatmap(x_src) if args.w_hpf > 0 else None
for i, y_trg in enumerate(y_trg_list):
z_many = torch.randn(10000, latent_dim).to(x_src.device)
y_many = torch.LongTensor(10000).to(x_src.device).fill_(y_trg[0])
s_many = nets.mapping_network(z_many, y_many)
s_avg = torch.mean(s_many, dim=0, keepdim=True)
s_avg = s_avg.repeat(N, 1)
for z_trg in z_trg_list:
s_trg = nets.mapping_network(z_trg, y_trg)
s_trg = torch.lerp(s_avg, s_trg, psi)
x_fake = nets.generator(x_src, s_trg, masks=masks)
x_concat += [x_fake]
x_concat = torch.cat(x_concat, dim=0)
save_image(x_concat, N, filename)
@torch.no_grad()
def translate_using_reference(nets, args, x_src, x_ref, y_ref, filename):
N, C, H, W = x_src.size()
wb = torch.ones(1, C, H, W).to(x_src.device)
x_src_with_wb = torch.cat([wb, x_src], dim=0)
masks = nets.fan.get_heatmap(x_src) if args.w_hpf > 0 else None
s_ref = nets.style_encoder(x_ref, y_ref)
s_ref_list = s_ref.unsqueeze(1).repeat(1, N, 1)
x_concat = [x_src_with_wb]
for i, s_ref in enumerate(s_ref_list):
x_fake = nets.generator(x_src, s_ref, masks=masks)
x_fake_with_ref = torch.cat([x_ref[i:i+1], x_fake], dim=0)
x_concat += [x_fake_with_ref]
x_concat = torch.cat(x_concat, dim=0)
save_image(x_concat, N+1, filename)
del x_concat
@torch.no_grad()
def debug_image(nets, args, inputs, step):
x_src, y_src = inputs.x_src, inputs.y_src
x_ref, y_ref = inputs.x_ref, inputs.y_ref
device = inputs.x_src.device
N = inputs.x_src.size(0)
# translate and reconstruct (reference-guided)
filename = ospj(args.sample_dir, '%06d_cycle_consistency.jpg' % (step))
translate_and_reconstruct(nets, args, x_src, y_src, x_ref, y_ref, filename)
# latent-guided image synthesis
y_trg_list = [torch.tensor(y).repeat(N).to(device)
for y in range(min(args.num_domains, 5))]
z_trg_list = torch.randn(args.num_outs_per_domain, 1, args.latent_dim).repeat(1, N, 1).to(device)
for psi in [0.5, 0.7, 1.0]:
filename = ospj(args.sample_dir, '%06d_latent_psi_%.1f.jpg' % (step, psi))
translate_using_latent(nets, args, x_src, y_trg_list, z_trg_list, psi, filename)
# reference-guided image synthesis
filename = ospj(args.sample_dir, '%06d_reference.jpg' % (step))
translate_using_reference(nets, args, x_src, x_ref, y_ref, filename)