-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinference.py
140 lines (129 loc) · 6.46 KB
/
inference.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
134
135
136
137
138
139
140
import argparse
import numpy as np
import torch
import math
import os
from tqdm import tqdm
from generators import generators
import configs
import numpy as np
import imageio
import cv2
def parse_seeds(seeds):
seeds = seeds.split(',')
res = []
for seed in seeds:
if '-' in seed:
start, end = seed.split('-')
res += list(range(int(start), int(end)+1))
else:
res.append(int(seed))
return res
def load_model(config, path):
generator_args = {}
if 'representation' in config['generator']:
generator_args['representation_kwargs'] = config['generator']['representation']['kwargs']
if 'super_resolution' in config['generator']:
generator_args['super_resolution_kwargs'] = config['generator']['super_resolution']['kwargs']
if 'renderer' in config['generator']:
generator_args['renderer_kwargs'] = config['generator']['renderer']['kwargs']
generator = getattr(generators, config['generator']['class'])(
**generator_args,
**config['generator']['kwargs']
)
generator.load_state_dict(torch.load(path, map_location='cpu'))
generator = generator.to('cuda')
generator.eval()
return generator
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--ckpt', type=str, default='./ckpts/ffhq1024.pt')
parser.add_argument('--output_dir', type=str, default='./images')
parser.add_argument('--config', type=str, default='GRAMHD1024_FFHQ')
parser.add_argument('--seeds', type=str, default='0-9')
parser.add_argument('--type', type=str, default='multiview1x3', choices=['video', 'multiview1x3', 'multiview3x7', 'rand'])
parser.add_argument('--truncation', type=float, default=0.7)
opt = parser.parse_args()
os.makedirs(opt.output_dir, exist_ok=True)
config = getattr(configs, opt.config)
generator = load_model(config, opt.ckpt)
# default setting
if opt.type == 'video':
h_mean = math.pi * (90 / 180)
v_mean = math.pi * (90 / 180)
generator.renderer.lock_view_dependence = True
frames = 60
concat = None
face_yaws = list(np.linspace(-0.3, 0.3, frames // 2 + 1)[:-1]) + list(np.linspace(0.3, -0.3, frames // 2 + 1)[:-1])
face_pitchs = [0*np.pi] * frames
face_angles = [[a + v_mean, b + h_mean] for a, b in zip(face_pitchs, face_yaws)]
fovs = [12] * frames
elif opt.type == 'multiview1x3':
h_mean = math.pi * (90 / 180)
v_mean = math.pi * (90 / 180)
generator.renderer.lock_view_dependence = True
frames = 3
concat = (1, 3)
face_yaws = list(np.linspace(-0.3, 0.3, frames))
face_pitchs = [0.0] * frames
face_angles = [[a + v_mean, b + h_mean] for a, b in zip(face_pitchs, face_yaws)]
fovs = [12] * frames
elif opt.type == 'multiview3x7':
h_mean = math.pi * (90 / 180)
v_mean = math.pi * (90 / 180)
generator.renderer.lock_view_dependence = True
frames = 21
concat = (3, 7)
face_yaws = list(np.linspace(-0.4, 0.4, frames//3)) * 3
face_pitchs = [0.2] * (frames // 3) + [0.0] * (frames // 3) + [-0.2] * (frames // 3)
face_angles = [[a + v_mean, b + h_mean] for a, b in zip(face_pitchs, face_yaws)]
fovs = [12] * frames
elif opt.type == 'rand':
generator.renderer.lock_view_dependence = False
frames = 1
concat = (1, 1)
face_angles = [[None, None]]
fovs = [12] * frames
seeds = parse_seeds(opt.seeds)
is_sr_model = isinstance(generator, generators.GramHDGenerator)
for idx, seed in tqdm(enumerate(seeds), total=len(seeds), desc='Generating images'):
images = np.zeros((frames, config['global']['img_size'], config['global']['img_size'], 3), dtype=np.uint8)
if is_sr_model:
lr_images = np.zeros((frames, config['global']['img_size'] // generator.scale_factor, config['global']['img_size'] // generator.scale_factor, 3), dtype=np.uint8)
for i, ((pitch, yaw), fov) in enumerate(zip(face_angles, fovs)):
config['camera']['fov'] = fov
torch.manual_seed(seed)
z = torch.randn((1, 256), device='cuda')
with torch.no_grad():
generator.get_avg_w()
camera_origin = [np.sin(pitch) * np.cos(yaw), np.cos(pitch), np.sin(pitch) * np.sin(yaw)] if pitch is not None and yaw is not None else None
img = generator(z, **config['camera'], camera_origin=camera_origin, truncation_psi=opt.truncation)[0]
if is_sr_model:
lr_img = img[2]
img = img[0]
lr_img = lr_img * 0.5 + 0.5
lr_img = lr_img.permute(0, 2, 3, 1).squeeze().cpu().numpy()
lr_img = (lr_img * 255).astype(np.uint8)
lr_images[i] = np.nan_to_num(lr_img)
img = img * 0.5 + 0.5
img = img.permute(0, 2, 3, 1).squeeze().cpu().numpy()
img = (np.clip(img, 0, 1) * 255).astype(np.uint8)
images[i] = np.nan_to_num(img)
if concat is None and opt.type != 'video':
imageio.imsave(os.path.join(opt.output_dir, f'grid_{seed}_{i}.png'),images[i])
if is_sr_model: imageio.imsave(os.path.join(opt.output_dir, f'grid_{seed}_{i}_lr.png'),lr_images[i])
if opt.type == 'video':
imageio.mimsave(os.path.join(opt.output_dir, f'grid_{seed}.mp4'), images, fps=30)
if is_sr_model: imageio.mimsave(os.path.join(opt.output_dir, f'grid_{seed}_lr.mp4'), lr_images, fps=30)
elif opt.type == 'rand':
imageio.imsave(os.path.join(opt.output_dir, f'{idx:05d}.png'), images[0])
elif concat is not None:
images = images.reshape((concat[0], concat[1], config['global']['img_size'], config['global']['img_size'], 3))
images = np.concatenate(images, axis=-3)
images = np.concatenate(images, axis=-2)
imageio.imsave(os.path.join(opt.output_dir, f'grid_{seed}.png'), images)
if is_sr_model:
lr_images = lr_images.reshape((concat[0], concat[1], config['global']['img_size'] // generator.scale_factor, config['global']['img_size'] // generator.scale_factor, 3))
lr_images = np.concatenate(lr_images, axis=-3)
lr_images = np.concatenate(lr_images, axis=-2)
imageio.imsave(os.path.join(opt.output_dir, f'grid_{seed}_lr.png'), lr_images)