This repository has been archived by the owner on May 4, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 135
/
Copy pathalign_images_video.py
166 lines (133 loc) · 5.99 KB
/
align_images_video.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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
import argparse
import cv2
import json
import numpy
from pathlib import Path
from tqdm import tqdm
from umeyama import umeyama
import queue
import threading
import av
from face_alignment import FaceAlignment, LandmarksType
FACE_ALIGNMENT = FaceAlignment( LandmarksType._2D, enable_cuda=True, flip_input=False )
mean_face_x = numpy.array([
0.000213256, 0.0752622, 0.18113, 0.29077, 0.393397, 0.586856, 0.689483, 0.799124,
0.904991, 0.98004, 0.490127, 0.490127, 0.490127, 0.490127, 0.36688, 0.426036,
0.490127, 0.554217, 0.613373, 0.121737, 0.187122, 0.265825, 0.334606, 0.260918,
0.182743, 0.645647, 0.714428, 0.793132, 0.858516, 0.79751, 0.719335, 0.254149,
0.340985, 0.428858, 0.490127, 0.551395, 0.639268, 0.726104, 0.642159, 0.556721,
0.490127, 0.423532, 0.338094, 0.290379, 0.428096, 0.490127, 0.552157, 0.689874,
0.553364, 0.490127, 0.42689 ])
mean_face_y = numpy.array([
0.106454, 0.038915, 0.0187482, 0.0344891, 0.0773906, 0.0773906, 0.0344891,
0.0187482, 0.038915, 0.106454, 0.203352, 0.307009, 0.409805, 0.515625, 0.587326,
0.609345, 0.628106, 0.609345, 0.587326, 0.216423, 0.178758, 0.179852, 0.231733,
0.245099, 0.244077, 0.231733, 0.179852, 0.178758, 0.216423, 0.244077, 0.245099,
0.780233, 0.745405, 0.727388, 0.742578, 0.727388, 0.745405, 0.780233, 0.864805,
0.902192, 0.909281, 0.902192, 0.864805, 0.784792, 0.778746, 0.785343, 0.778746,
0.784792, 0.824182, 0.831803, 0.824182 ])
landmarks_2D = numpy.stack( [ mean_face_x, mean_face_y ], axis=1 )
def transform( image, mat, size, padding=0 ):
mat = mat * size
mat[:,2] += padding
new_size = int( size + padding * 2 )
return cv2.warpAffine( image, mat, ( new_size, new_size ) )
def prepare_image(input_dir, seekstart, durationtime, fps, workerqueue):
container = av.open(input_dir)
stream = container.streams.video[0]
container.seek(seekstart*1000000)
frame = next(f for f in container.decode(video=0))
videostart = frame.pts * stream.time_base
endtime = seekstart + durationtime
d = min(endtime, stream.duration * stream.time_base) - videostart
pbar = tqdm(total=(int(d*min(stream.average_rate, fps))+1), ascii=True)
image = frame.to_nd_array(format='rgb24')
workerqueue.put((frame.pts, image))
pbar.update(1)
fps = 1/fps
timenext = videostart + fps
for frame in container.decode(video=0):
timenow = frame.pts * stream.time_base
if timenow > endtime:
break
if (timenow >= timenext):
timenext += fps
image = frame.to_nd_array(format='rgb24')
workerqueue.put((frame.pts, image))
pbar.update(1)
workerqueue.put(None)
pbar.close()
def find_face(inputqueue, outputqueue):
while True:
item = inputqueue.get()
if item is None:
break
idx, image = item
faces = FACE_ALIGNMENT.get_landmarks( image )
if faces is None: continue
if len(faces) == 0: continue
outputqueue.put((idx, image, faces))
outputqueue.put(None)
def align_face(output_dir, workerqueue):
while True:
item = workerqueue.get()
if item is None:
break
idx, image, faces = item
image = image[:, :, ::-1]
for i,points in enumerate(faces):
alignment = umeyama( points[17:], landmarks_2D, True )[0:2]
aligned_image = transform( image, alignment, 80, 24 )
# aligned_image = transform( image, alignment, 160, 48 )
if len(faces) == 1:
out_fn = "{:0>10}.jpg".format( idx )
else:
out_fn = "{:0>10}_{}.jpg".format( idx, i )
out_fn = output_dir / out_fn
cv2.imwrite( str(out_fn), aligned_image )
yield idx, str(out_fn.relative_to(output_dir)), list( alignment.ravel() ), list(points.flatten().astype(float))
def main( args ):
output_dir = Path(args.output_dir)
output_dir.mkdir( parents=True, exist_ok=True )
output_file = output_dir / args.alignments
queue_prepare_image = queue.Queue(maxsize=1)
thread_prepare_image = threading.Thread(target=prepare_image, args=(args.input_dir, args.seekstart, args.durationtime, args.fps, queue_prepare_image))
thread_prepare_image.start()
queue_find_face = queue.Queue(maxsize=1)
thread_find_face = threading.Thread(target=find_face, args=(queue_prepare_image, queue_find_face))
thread_find_face.start()
face_alignments = list( align_face(output_dir, queue_find_face) )
with output_file.open('w') as f:
results = json.dumps( face_alignments, ensure_ascii=False )
f.write( results )
print( "Save face alignments to output file:", output_file )
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('-i', '--input-dir',
dest="input_dir",
type=str,
default="input.mp4",
help="Input video")
parser.add_argument('-o', '--output-dir',
dest="output_dir",
type=str,
default="aligned",
help="Output directory. This is where the aligned images will \
be stored. Defaults to 'aligned'")
parser.add_argument('-a', '--alignments',
dest="alignments",
type=str,
default='alignments.json')
parser.add_argument('-ss', '--seekstart',
dest="seekstart",
type=int,
default=0)
parser.add_argument('-t', '--durationtime',
dest="durationtime",
type=int,
default=100000)
parser.add_argument('-vf', '--fps',
dest="fps",
type=float,
default=1000.0)
main( parser.parse_args() )