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
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
|
#!/usr/bin/env python2.7
import os
import sys
import getopt
import numpy
import torch
import PIL
import PIL.Image
from datetime import datetime
from network import Network, process
from moviepy.video.io.ffmpeg_reader import FFMPEG_VideoReader
from moviepy.video.io.ffmpeg_writer import FFMPEG_VideoWriter
##########################################################
arguments_strModel = 'lf'
arguments_strFirst = './images/first.png'
arguments_strSecond = './images/second.png'
arguments_strOut = './result.png'
arguments_strVideo = False
arguments_strVideoOut = datetime.now().strftime("sepconv_%Y%m%d_%H%M.mp4")
arguments_steps = 0
arguments_dilate = 1
arguments_aOffset = 0
arguments_bOffset = 0
arguments_mixVideos = False
for strOption, strArgument in getopt.getopt(sys.argv[1:], '', [ strParameter[2:] + '=' for strParameter in sys.argv[1::2] ])[0]:
print("{}: {}".format(strOption, strArgument))
if strOption == '--model':
arguments_strModel = strArgument # which model to use, l1 or lf, please see our paper for more details
elif strOption == '--first':
arguments_strFirst = strArgument # path to the first frame
elif strOption == '--second':
arguments_strSecond = strArgument # path to the second frame
elif strOption == '--out':
arguments_strOut = strArgument # path to where the output should be stored
elif strOption == '--video':
arguments_strVideoOut = strArgument # path to video
elif strOption == '--video-out':
arguments_strVideoOut = strArgument # path to where the video should be stored
elif strOption == '--steps':
arguments_steps = int(strArgument)
elif strOption == '--dilate':
arguments_dilate = int(strArgument)
elif strOption == '--mix-videos':
arguments_mixVideos = True
elif strOption == '--a-offset':
arguments_aOffset = int(strArgument)
elif strOption == '--b-offset':
arguments_bOffset = int(strArgument)
if not os.path.exists('./renders'):
os.mkdir('renders')
moduleNetwork = Network(arguments_strModel).cuda()
tensorOutput = torch.FloatTensor()
index = 0
def recurse_two_frames(moduleNetwork, tensorOutput, a_np, b_np, frame_index, morph_index, step, depth=0):
print("generate {} {} {}".format(frame_index, morph_index, step))
tensorInputFirst = torch.FloatTensor(a_np)
tensorInputSecond = torch.FloatTensor(b_np)
process(moduleNetwork, tensorInputFirst, tensorInputSecond, tensorOutput)
middle_np = tensorOutput.clamp(0.0, 1.0).numpy()
if morph_index == frame_index:
print("frame {}, depth {}".format(frame_index, depth))
return middle_np
if morph_index > frame_index:
next_index = morph_index - step
next_a_np = a_np
next_b_np = middle_np
# print("next index: {} - {}".format(next_index, step))
else:
next_index = morph_index + step
next_a_np = middle_np
next_b_np = b_np
# print("next index: {} + {}".format(next_index, step))
return recurse_two_frames(moduleNetwork, tensorOutput, next_a_np, next_b_np, frame_index, next_index, step/2, depth+1)
def recurse_videos(moduleNetwork, tensorOutput, a, b, a_offset, b_offset, count, step, frame_index, dilate):
global index
index += 1
if (index % 10) == 0:
print("{}...".format(index))
step /= 2
a_fn = os.path.join(a, "frame_{:0>5}.png".format(int(frame_index + a_offset)))
b_fn = os.path.join(b, "frame_{:0>5}.png".format(int(frame_index + b_offset)))
print("{} => {}".format(a_fn, b_fn))
a_np = load_image(a_fn)
b_np = load_image(b_fn)
img_np = recurse_two_frames(moduleNetwork, tensorOutput, a_np, b_np, frame_index, count / 2, count / 4)
if step < 2 * dilate:
return [img_np]
else:
left = recurse_videos(moduleNetwork, tensorOutput, a, b, a_offset, b_offset, count, step, frame_index - (step/2), dilate)
right = recurse_videos(moduleNetwork, tensorOutput, a, b, a_offset, b_offset, count, step, frame_index + (step/2), dilate)
return left + [img_np] + right
def process_two_videos(moduleNetwork, tensorOutput, a, b, a_offset, b_offset, steps, dilate):
steps *= dilate
return recurse_videos(moduleNetwork, tensorOutput, a, b, a_offset, b_offset, steps, steps, steps/2, dilate)
def process_tree(moduleNetwork, a, b, tensorOutput, steps, dilate):
global index
index += 1
if (index % 10) == 0:
print("{}...".format(index))
tensorInputFirst = torch.FloatTensor(a)
tensorInputSecond = torch.FloatTensor(b)
process(moduleNetwork, tensorInputFirst, tensorInputSecond, tensorOutput)
middle_np = tensorOutput.clamp(0.0, 1.0).numpy()
if steps < 4 * dilate:
return [middle_np]
else:
left = process_tree(moduleNetwork, a, middle_np, tensorOutput, steps / 2, dilate)
right = process_tree(moduleNetwork, middle_np, b, tensorOutput, steps / 2, dilate)
return left + [middle_np] + right
def dilate_frames(moduleNetwork, tensorOutput, frames, dilate):
if dilate < 2:
return frames
print("dilating by a factor of {}".format(dilate))
new_frames = []
nextFrame = frames[0]
for i in range(1, len(frames)):
firstFrame = nextFrame
nextFrame = frames[i]
new_frames += [firstFrame]
new_frames += process_tree(moduleNetwork, firstFrame, nextFrame, tensorOutput, dilate, 1)
new_frames += nextFrame
return new_frames
def store_frames(frames, outputPath, inputFirst=None, inputSecond=None):
print('writing {}'.format(outputPath))
print('frames: {}'.format(len(frames)))
writer = FFMPEG_VideoWriter(outputPath, (1024, 512), 25)
if inputFirst is not None:
writer.write_frame(inputFirst)
for frame in frames:
writer.write_frame(tensor_to_image(frame))
if inputSecond is not None:
writer.write_frame(inputSecond)
def tensor_to_image(np_val):
return (numpy.rollaxis(np_val, 0, 3)[:,:,::-1] * 255.0).astype(numpy.uint8)
def load_image(path):
return numpy.rollaxis(numpy.asarray(PIL.Image.open(path))[:,:,::-1], 2, 0).astype(numpy.float32) / 255.0
def load_image_tensor(path):
return torch.FloatTensor(load_image(path))
if arguments_strVideo and arguments_strVideoOut:
reader = FFMPEG_VideoReader(arguments_strVideo, False)
writer = FFMPEG_VideoWriter(arguments_strVideoOut, reader.size, reader.fps*2)
reader.initialize()
nextFrame = reader.read_frame()
for x in range(0, reader.nframes):
firstFrame = nextFrame
nextFrame = reader.read_frame()
tensorInputFirst = torch.FloatTensor(numpy.rollaxis(firstFrame[:,:,::-1], 2, 0) / 255.0)
tensorInputSecond = torch.FloatTensor(numpy.rollaxis(nextFrame[:,:,::-1], 2, 0) / 255.0)
process(moduleNetwork, tensorInputFirst, tensorInputSecond, tensorOutput)
writer.write_frame(firstFrame)
writer.write_frame((numpy.rollaxis(tensorOutput.clamp(0.0, 1.0).numpy(), 0, 3)[:,:,::-1] * 255.0).astype(numpy.uint8))
writer.write_frame(nextFrame)
writer.close()
elif arguments_mixVideos:
# Morph two videos
print("morph two videos...")
print("{} => {}".format(arguments_strFirst, arguments_strSecond))
outputPath = './renders/' + arguments_strVideoOut
frames = process_two_videos(moduleNetwork, tensorOutput, arguments_strFirst, arguments_strSecond, arguments_aOffset, arguments_bOffset, arguments_steps, arguments_dilate)
dilate_frames(moduleNetwork, tensorOutput, frames, arguments_dilate)
store_frames(frames, outputPath)
elif arguments_steps == 0:
# Process image
tensorInputFirst = load_image_tensor(arguments_strFirst)
tensorInputSecond = load_image_tensor(arguments_strSecond)
process(moduleNetwork, tensorInputFirst, tensorInputSecond, tensorOutput)
PIL.Image.fromarray((numpy.rollaxis(tensorOutput.clamp(0.0, 1.0).numpy(), 0, 3)[:,:,::-1] * 255.0).astype(numpy.uint8)).save(arguments_strOut)
else:
# Morph two images
print("{} => {}".format(arguments_strFirst, arguments_strSecond))
inputFirst = load_image(arguments_strFirst)
inputSecond = load_image(arguments_strSecond)
outputPath = './renders/' + arguments_strVideoOut
frames = process_tree(moduleNetwork, inputFirst, inputSecond, tensorOutput, arguments_steps * arguments_dilate, arguments_dilate)
dilate_frames(moduleNetwork, tensorOutput, frames, arguments_dilate)
store_frames(frames, outputPath, inputFirst, inputSecond)
|
