-
Notifications
You must be signed in to change notification settings - Fork 3
/
inference.py
84 lines (72 loc) · 2.8 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
import torch
import cv2
from matplotlib import pyplot as plt
from network.model import Generator
from video_extraction import generate_landmarks
import os.path
def main(model_weights_path: str = 'model_weights.tar',
embedding_path: str = 'e_hat_video.tar',
video_path: str = 'examples/fine_tuning/test_video.mp4',
output_dir: str = './'):
"""Init"""
device = torch.device("cuda:0")
cpu = torch.device("cpu")
checkpoint = torch.load(model_weights_path, map_location=cpu)
e_hat = torch.load(embedding_path, map_location=cpu)
e_hat = e_hat['e_hat'].to(device)
generator = Generator(256, finetuning=True, e_finetuning=e_hat)
generator.eval()
"""Training Init"""
generator.load_state_dict(checkpoint['G_state_dict'])
generator.to(device)
generator.finetuning_init()
"""Main"""
print('PRESS Q TO EXIT')
cap = cv2.VideoCapture(video_path)
with torch.no_grad():
enum = 0
while True:
print("doing enum", enum)
x, g_y = generate_landmarks(cap=cap, device=device, pad=50)
if x is None and g_y is None:
print("broke at enum ", enum)
break
g_y = g_y.unsqueeze(0)
x = x.unsqueeze(0)
# forward
# Calculate average encoding vector for video
# f_lm_compact = f_lm.view(-1, f_lm.shape[-4], f_lm.shape[-3], f_lm.shape[-2], f_lm.shape[-1]) #BxK,2,3,224,224
# train generator
x_hat = generator(g_y, e_hat)
plt.clf()
out1 = x_hat.transpose(1, 3)[0] / 255
for img_no in range(1,x_hat.shape[0]):
out1 = torch.cat((out1, x_hat.transpose(1,3)[img_no]), dim = 1)
out1 = out1.to(cpu).numpy()
plt.imshow(out1)
plt.show()
plt.imsave(os.path.join(output_dir, 'fake-{}.png'.format(enum)), out1)
plt.clf()
out2 = x.transpose(1, 3)[0] / 255
for img_no in range(1,x.shape[0]):
out2 = torch.cat((out2, x.transpose(1,3)[img_no]), dim = 1)
out2 = out2.to(cpu).numpy()
plt.imshow(out2)
plt.show()
plt.imsave(os.path.join(output_dir, 'head_track-{}.png'.format(enum)), out2)
plt.clf()
out3 = g_y.transpose(1, 3)[0] / 255
for img_no in range(1,g_y.shape[0]):
out3 = torch.cat((out3, g_y.transpose(1,3)[img_no]), dim = 1)
out3 = out3.to(cpu).numpy()
plt.imshow(out3)
plt.show()
plt.imsave(os.path.join(output_dir, 'landmark-{}.png'.format(enum)), out3)
plt.clf()
if cv2.waitKey(1) == ord('q'):
break
enum += 1
cap.release()
cv2.destroyAllWindows()
if __name__ == "__main__":
main()