-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmmvc_dataset.py
212 lines (191 loc) · 8.27 KB
/
mmvc_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
212
'''
参考
https://github.com/BowenL0218/MMVC_video_codec/blob/main/datasets.py
'''
import torch
import torchvision
from torch import nn
from torch.autograd import Variable
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from torchvision.utils import save_image
import PIL
import os
class VimeoDataset(Dataset):
def __init__(self, video_dir, text_split, test=False, transform=None):
"""
Dataset class for the Vimeo-90k dataset, available at http://toflow.csail.mit.edu/.
Args:
video_dir (string): Vimeo-90k sequences directory.
text_split (string): Text file path in the Vimeo-90k folder, either `tri_trainlist.txt` or `tri_testlist.txt`.
transform (callable, optional): Optional transform to be applied samples.
"""
self.video_dir = video_dir
self.test = test
self.text_split = text_split
# default transform as per RRIN, convert images to tensors, with values between 0 and 1
if transform is None:# and test is False:
self.transform = transforms.Compose([
# transforms.RandomCrop((256, 256)),
transforms.ToTensor(), # ToTensor() 操作包含scale到[0, 1]和HWC to CHW
])
else:
self.transform = transforms.Compose([
transforms.ToTensor(),
])
# self.prev1_frame = []
# self.prev2_frame = []
# self.prev3_frame = []
# self.prev4_frame = []
self.ref_frame = []
self.cur_frame = []
# open the given text file path that gives file names for train or test subsets
with open(self.text_split, 'r') as f:
filenames = f.readlines()
f.close()
full_filenames = []
for i in filenames:
full_filenames.append(os.path.join(self.video_dir, i.split('\n')[0]))
for f in full_filenames:
try:
frames = [os.path.join(f, i) for i in os.listdir(f)]
except:
continue
# make sure images are in order, i.e. im1.png, im2.png, im3.png
frames = sorted(frames)
# make sure there are only 3 images in the Vimeo-90k triplet's folder for it to be a valid dataset sample
# if len(frames) == 3:
# self.prev1_frame.append(frames[0])
# self.prev2_frame.append(frames[1])
# self.cur_frame.append(frames[2])
# if len(frames) == 7:
# for i in range(3):
# self.prev1_frame.append(frames[i])
# self.prev2_frame.append(frames[i+1])
# self.prev3_frame.append(frames[i+2])
# self.prev4_frame.append(frames[i+3])
# self.cur_frame.append(frames[i+4])
# print(frames)
if len(frames) == 3:
for i in range(1):
self.ref_frame.append(frames[i])
self.cur_frame.append(frames[i+1])
if len(frames) == 7:
for i in range(6):
self.ref_frame.append(frames[i])
self.cur_frame.append(frames[i+1])
print(f"Dataset loaded with {len(self.cur_frame)} samples.")
def __len__(self):
return len(self.cur_frame)
def __getitem__(self, idx):
# prev1 = PIL.Image.open(self.prev1_frame[idx]).convert("RGB")
# prev2 = PIL.Image.open(self.prev2_frame[idx]).convert("RGB")
# prev3 = PIL.Image.open(self.prev3_frame[idx]).convert("RGB")
# prev4 = PIL.Image.open(self.prev4_frame[idx]).convert("RGB")
# cur = PIL.Image.open(self.cur_frame[idx]).convert("RGB")
ref = PIL.Image.open(self.ref_frame[idx]).convert("RGB")
cur = PIL.Image.open(self.cur_frame[idx]).convert("RGB")
if self.transform:
# prev1 = self.transform(prev1)
# prev2 = self.transform(prev2)
# prev3 = self.transform(prev3)
# prev4 = self.transform(prev4)
# cur = self.transform(cur)
ref = self.transform(ref)
cur = self.transform(cur)
if self.test is False:
# Concat = torch.cat([prev1, prev2, cur], axis = 0)
# Concat = torch.cat([prev1, prev2, prev3, prev4, cur], axis = 0)
# print(Concat.shape)
Concat = torch.cat([ref, cur], axis = 0)
transform = transforms.Compose([
transforms.RandomCrop((256, 256)),
# transforms.ToTensor(),
])
Concat = transform(Concat)
# prev1 = Concat[:3,:,:]
# prev2 = Concat[3:6,:,:]
# prev3 = Concat[6:9,:,:]
# prev4 = Concat[9:12,:,:]
# cur = Concat[12:,:,:]
ref_image = Concat[:3,:,:]
input_image = Concat[3:,:,:]
else:
input_image = cur
ref_image = ref
return input_image, ref_image
class VimeoDatasetEx(Dataset):
def __init__(self, video_dir, text_split, ref_num, test=False, transform=None):
"""
Dataset class for the Vimeo-90k dataset, available at http://toflow.csail.mit.edu/.
Args:
video_dir (string): Vimeo-90k sequences directory.
text_split (string): Text file path in the Vimeo-90k folder, either `tri_trainlist.txt` or `tri_testlist.txt`.
transform (callable, optional): Optional transform to be applied samples.
"""
self.video_dir = video_dir
self.test = test
self.text_split = text_split
# default transform as per RRIN, convert images to tensors, with values between 0 and 1
if transform is None:# and test is False:
self.transform = transforms.Compose([
# transforms.RandomCrop((256, 256)),
transforms.ToTensor(), # ToTensor() 操作包含scale到[0, 1]和HWC to CHW
])
else:
self.transform = transforms.Compose([
transforms.ToTensor(),
])
self.ref_frame = [[] for i in range(ref_num)]
self.cur_frame = []
# open the given text file path that gives file names for train or test subsets
with open(self.text_split, 'r') as f:
filenames = f.readlines()
f.close()
full_filenames = []
for i in filenames:
full_filenames.append(os.path.join(self.video_dir, i.split('\n')[0]))
for f in full_filenames:
try:
frames = [os.path.join(f, i) for i in os.listdir(f)]
except:
continue
# make sure images are in order, i.e. im1.png, im2.png, im3.png
frames = sorted(frames)
# if len(frames) == 3:
# for i in range(1):
# self.ref_frame.append(frames[i])
# self.cur_frame.append(frames[i+1])
if len(frames) == 7:
for i in range(7-ref_num):
for j in range(ref_num):
self.ref_frame[j].append(frames[i+j])
self.cur_frame.append(frames[i+ref_num])
print(f"Dataset loaded with {len(self.cur_frame)} samples.")
def __len__(self):
return len(self.cur_frame)
def __getitem__(self, idx):
ref = []
for i in range(len(self.ref_frame)):
ref.append(PIL.Image.open(self.ref_frame[i][idx]).convert("RGB"))
cur = PIL.Image.open(self.cur_frame[idx]).convert("RGB")
if self.transform:
for i in range(len(ref)):
ref[i] = self.transform(ref[i])
cur = self.transform(cur)
if self.test is False:
Concat = torch.cat(ref, axis = 0)
Concat = torch.cat([Concat, cur], axis = 0)
transform = transforms.Compose([
transforms.RandomCrop((256, 256)),
# transforms.ToTensor(),
])
Concat = transform(Concat)
ref_image_list = []
for i in range(len(self.ref_frame)):
ref_image_list.append(Concat[3*i:3*(i+1),:,:])
cur_image = Concat[3*len(self.ref_frame):,:,:]
else:
cur_image = cur
ref_image_list = ref
return cur_image, ref_image_list