-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathmake_fakevideo_dataset.py
212 lines (174 loc) · 8.29 KB
/
make_fakevideo_dataset.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
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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
#pip install moviepy
import argparse
from PIL import Image
from torchvision.utils import save_image, make_grid
import torchvision.transforms as tfs
import torchvision.transforms.functional as F
from moviepy.editor import *
import torch
import os
import random
import math
import multiprocessing
from joblib import Parallel,delayed
import time
import random
from PIL import ImageFilter
import random
import numpy as np
import matplotlib.pyplot as plt
from pytorchvideo.transforms import AugMix
from moviepy.editor import ImageSequenceClip
class MyRandomResizedCrop(tfs.RandomResizedCrop):
@staticmethod
def get_params(
img, scale, ratio):
"""Get parameters for ``crop`` for a random sized crop.
# compared to original: there's no fallback of doing a CentralCrop
Args:
img (PIL Image or Tensor): Input image.
scale (list): range of scale of the origin size cropped
ratio (list): range of aspect ratio of the origin aspect ratio cropped
Returns:
tuple: params (i, j, h, w) to be passed to ``crop`` for a random
sized crop.
"""
width, height = img.size[0], img.size[1]
area = height * width
while True: # basically no falling back to center-crop, as in the original implementation
target_area = area * (
torch.empty(1).uniform_(scale[0], scale[1]).item() # **2 # the cubed is new
)
log_ratio = torch.log(torch.tensor(ratio))
aspect_ratio = torch.exp(
torch.empty(1).uniform_(log_ratio[0], log_ratio[1])
).item()
w = int(round(math.sqrt(target_area * aspect_ratio)))
h = int(round(math.sqrt(target_area / aspect_ratio)))
if 0 < w <= width and 0 < h <= height:
i = torch.randint(0, height - h + 1, size=(1,)).item()
j = torch.randint(0, width - w + 1, size=(1,)).item()
return i, j, h, w
class MonoDataset(torch.utils.data.Dataset):
def __init__(self, image_path, initcrop=0.5, scale=(0.01, 1.0), vflip=True, hflip=True,
crop_size=256, mean=[0, 0, 0], std=[1, 1, 1], cropfirst=True, frames=8,
length=int(1.281e6)):
self.frames = frames
self.img = None
self.image_path = image_path
normalize = tfs.Normalize(mean=mean, std=std)
self.img = (Image.open(self.image_path).convert('RGB'))
self.scale = (1./scale[0], 1./scale[1])
self.resizedcrop = MyRandomResizedCrop(int(math.sqrt(2) * crop_size), scale=self.scale, interpolation='bicubic')
self.length = length # run out after one epoch
self.crop_size = crop_size
self.augmix = AugMix()
if cropfirst:
self.init_crop = tfs.RandomCrop(int(initcrop*min(self.img.size)))
if vflip:
self.vflip = tfs.RandomVerticalFlip(p=0.5)
if hflip:
self.hflip = tfs.RandomHorizontalFlip(p=0.5)
tfslist = []
tfslist.append(tfs.CenterCrop(self.crop_size))
self.cj = tfs.RandomApply([tfs.ColorJitter(brightness=0.4, contrast=0.4, saturation=0.4, hue=0.1)], 0.5)
tfslist.append(tfs.ToTensor())
if mean != [0, 0, 0]:
tfslist.append(normalize)
self.transforms = tfs.Compose(tfslist)
def __len__(self):
return self.length
def __getitem__(self, idx):
params = []
if self.img is None:
self.img = Image.open(self.image_path).convert('RGB')
img = self.init_crop(self.img)
img = self.cj(img)
steps = 2
for _ in range(steps):
params.append(self.resizedcrop.get_params(img, scale=self.scale, ratio=(3. / 4., 4. / 3.)))
i,j,h,w = params[0]
i2,j2,h2,w2 = params[1]
i_frames, j_frames = np.linspace(i,i2,self.frames),np.linspace(j,j2,self.frames)
h_frames, w_frames = np.linspace(h,h2,self.frames),np.linspace(w,w2,self.frames)
frames = []
for f in range(self.frames):
frames.append(F.resized_crop(img, i_frames[f], j_frames[f], h_frames[f], w_frames[f],
(self.crop_size, self.crop_size), Image.BICUBIC))
return torch.cat([self.transforms(f).unsqueeze(0) for f in frames],dim=0)
def get_mono_data_loader(batch_size, num_workers, **kwargs):
m = MonoDataset(**kwargs)
return torch.utils.data.DataLoader(m, batch_size=batch_size, num_workers=num_workers,
drop_last=True)
def save_batch_vids(imgs, count, pth, thread):
all_files = []
for k in range(imgs.size(0)):
files = []
for frame in range(imgs.size(1)):
name = pth+f'/_tmp_patch_{k+count}_f{frame}.png'
save_image(imgs[k,frame,:,:], name, padding=0,
normalize=False, scale_each=False)
files.append(name)
all_files.append(name)
clip = ImageSequenceClip(files, fps = 12)
clip.write_videofile(pth+ f"clips/clip_{thread}_{k+count}.mp4", fps = 15)
for f in all_files:
os.remove(f)
def make_(thred, args):
random.seed(555*thred)
torch.manual_seed(555*thred)
data_loader = get_mono_data_loader(args.batch_size, 0,
image_path=args.imgpath,
initcrop=args.initcrop,
scale=tuple(args.scale),
vflip=not args.no_vflip,
crop_size=args.img_size,
cropfirst=args.cropfirst, length=args.img_per_thread,
)
count = thred*args.img_per_thread
print(f"{len(data_loader.dataset) + (args.img_per_thread % args.batch_size)}images per thread")
for patches in data_loader:
count += len(patches)*args.frames
save_batch_vids(patches, count, path,thred)
# last batch: treat specially
diter = iter(data_loader)
patches = next(diter)[:(args.img_per_thread % args.batch_size)]
count += len(patches)
save_batch_vids(patches, count, path, thred)
return 0
def get_parser():
parser = argparse.ArgumentParser(description='Single Image Pretraining, Asano et al. 2020')
# Generation settings
parser.add_argument('--img_size', default=230, type=int, help='Size of generated images (default:256)')
parser.add_argument('--batch_size', default=50, type=int, help='Batchsize for generation (default:32)')
parser.add_argument('--num_imgs', default=200000, type=int, help='Number of images to be generated (default: 1281167)')
parser.add_argument('--threads', default=20, type=int, help='how many CPU threads to use for generation (default: 20)')
# Flipping
parser.add_argument('--no_vflip', action='store_false', help='use vflip? (default: True)')
# Scale/Crop augmentations
parser.add_argument('--cropfirst', action='store_false', help='usage of initial crop to not focus too much on center (default: True)')
parser.add_argument('--initcrop', default=0.5, type=float, help='initial crop size relative to image (default: 0.5)')
parser.add_argument('--scale', default=[500, 1], nargs=2, type=float, help='data augmentation inverse scale (default: 500, 1)')
# video stuff
parser.add_argument('--frames', default=16, type=int, help='number of frames') # fps is fixed to 12
# storing etc.
parser.add_argument('--imgpath', default="images/img_b.jpg", type=str)
parser.add_argument('--targetpath', default="/scratch/local/nvme/yuki/data/kd/", type=str)
return parser
if __name__ == "__main__":
args = get_parser().parse_args()
args.img_per_thread = args.num_imgs//args.threads
random.seed(555)
torch.manual_seed(555)
path = args.targetpath +"/" + str(args.img_size) + "_single" + args.imgpath.split('/')[-1].split('.')[0]
path += f"_init{args.initcrop}_scale{args.scale[0]}_{args.scale[1]}"
path += f"_novflip{args.no_vflip}_cropfirst{args.cropfirst}_new_{args.num_imgs}_noaugmix_frames_{args.frames}"
path += '/train/'
os.makedirs(path+"clips", exist_ok=True)
args.img_per_thread = (args.num_imgs//args.threads)
time.sleep(1)
t0 = time.time()
q = Parallel(n_jobs=args.threads)(
delayed(make_)(i,args)
for i in range(args.threads)
)