-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdataset_sketchy.py
193 lines (163 loc) · 8.11 KB
/
dataset_sketchy.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
import random
import numpy as np
import torch
import torchvision.transforms as transforms
import torch.utils.data as data
import torchvision.transforms.functional as F
import argparse
import pickle
import os
import time
from random import randint
from PIL import Image
import torchvision
import functools
# from render_sketch_chairv2 import redraw_Quick2RGB
# def get_ransform(opt):
# transform_list = []
# if opt.Train:
# transform_list.extend([transforms.Resize(320), transforms.CenterCrop(299)])
# else:
# transform_list.extend([transforms.Resize(299)])
# transform_list.extend(
# [transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
# return transforms.Compose(transform_list)
def get_ransform(opt):
transform_list = []
if opt.Train:
transform_list.extend([
transforms.RandomRotation((-5,5), resample=2)])
transform_list.extend([transforms.Resize(32), transforms.CenterCrop(28)])
else:
transform_list.extend([transforms.Resize(28)])
transform_list.extend(
# [transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
[transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))])
return transforms.Compose(transform_list)
def compare(a, b):
if int(a.split('_')[-1].split('-')[0]) < int(b.split('_')[-1].split('-')[0]):
return -1
elif int(a.split('_')[-1].split('-')[0]) == int(b.split('_')[-1].split('-')[0]):
if int(a.split('_')[-1].split('-')[-1].split('.')[0]) < int(b.split('_')[-1].split('-')[-1].split('.')[0]):
return -1
else:
return 1
else:
return 1
class CreateDataset_Sketchy(data.Dataset):
def __init__(self, opt, on_Fly=False):
# with open(opt.coordinate, 'rb') as fp:
# self.Coordinate = pickle.load(fp)
#
# self.Skecth_Train_List = [x for x in self.Coordinate if 'train' in x]
# self.Skecth_Test_List = [x for x in self.Coordinate if 'test' in x]
filenames = os.listdir(os.path.join(opt.roor_dir,'sketch','airplane'))
filenames.sort(key=functools.cmp_to_key(compare))
# print(filenames)
self.Skecth_Train_List = filenames[:501]
self.Skecth_Test_List = filenames[501:]
self.opt = opt
self.transform = get_ransform(opt)
self.on_Fly = on_Fly
def __getitem__(self, item):
if self.opt.mode == 'Train':
sketch_path = self.Skecth_Train_List[item]
sketch_signature = sketch_path.split('-')[0].split('_')[-1]
positive_sample = self.Skecth_Train_List[item].split('-')[0]
positive_path = os.path.join(self.opt.roor_dir, 'image', 'airplane', positive_sample + '.jpg')
possible_list = list(range(len(self.Skecth_Train_List)))
possible_list.remove(item)
flag = True
while(flag):
negetive_item = possible_list[randint(0, len(possible_list) - 1)]
negetive_prefix = self.Skecth_Train_List[negetive_item].split('-')[0].split('_')[-1]
if(negetive_prefix!=sketch_signature):
flag = False
negetive_sample = self.Skecth_Train_List[negetive_item].split('-')[0]
negetive_path = os.path.join(self.opt.roor_dir, 'image', 'airplane', negetive_sample + '.jpg')
sketch_img = []
sketch_img.append(Image.open(os.path.join(self.opt.roor_dir, 'sketch', 'airplane', sketch_path)))
# sketch_img[-1].show()
# if self.on_Fly == False:
# sketch_img = Image.fromarray(sketch_img[-1]).convert('RGB')
# else:
# sketch_img = [Image.fromarray(sk_img).convert('RGB') for sk_img in sketch_img]
positive_img = Image.open(positive_path)
negetive_img = Image.open(negetive_path)
n_flip = random.random()
sketch_img = sketch_img[-1].convert('L')
if n_flip > 0.5:
if self.on_Fly == False:
sketch_img = F.hflip(sketch_img)
else:
sketch_img = [F.hflip(sk_img) for sk_img in sketch_img]
positive_img = F.hflip(positive_img)
negetive_img = F.hflip(negetive_img)
if self.on_Fly == False:
sketch_img = self.transform(sketch_img)
else:
sketch_img = [self.transform(sk_img) for sk_img in sketch_img]
positive_img = self.transform(positive_img)
negetive_img = self.transform(negetive_img)
sample = {'sketch_img': sketch_img, 'sketch_path': self.Skecth_Train_List[item],
'positive_img': positive_img, 'positive_path': positive_sample,
'negetive_img': negetive_img, 'negetive_path': negetive_sample,
}
elif self.opt.mode == 'Test':
sketch_path = self.Skecth_Test_List[item]
sketch_signature = sketch_path.split('-')[0].split('_')[-1]
positive_sample = self.Skecth_Test_List[item].split('-')[0]
positive_path = os.path.join(self.opt.roor_dir, 'image', 'airplane', positive_sample + '.jpg')
possible_list = list(range(len(self.Skecth_Test_List)))
possible_list.remove(item)
flag = True
while (flag):
negetive_item = possible_list[randint(0, len(possible_list) - 1)]
negetive_prefix = self.Skecth_Train_List[negetive_item].split('-')[0].split('_')[-1]
if (negetive_prefix != sketch_signature):
flag = False
negetive_sample = self.Skecth_Train_List[negetive_item].split('-')[0]
negetive_path = os.path.join(self.opt.roor_dir, 'image', 'airplane', negetive_sample + '.jpg')
sketch_img = []
sketch_img.append(Image.open(os.path.join(self.opt.roor_dir, 'sketch', 'airplane', sketch_path)))
sketch_img = sketch_img[-1].convert('L')
if self.on_Fly == False:
sketch_img = self.transform(sketch_img)
else:
sketch_img = [self.transform(Image.fromarray(sk_img).convert('RGB')) for sk_img in sketch_img]
positive_img = self.transform(Image.open(positive_path))
negetive_img = self.transform(Image.open(negetive_path))
sample = {'sketch_img': sketch_img, 'sketch_path': self.Skecth_Test_List[item],
'positive_img': positive_img,
'negetive_img': negetive_img, 'negetive_path': negetive_sample,
'positive_path': positive_sample}
return sample
def __len__(self):
if self.opt.mode == 'Train':
return len(self.Skecth_Train_List)
elif self.opt.mode == 'Test':
return len(self.Skecth_Test_List)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
opt = parser.parse_args()
# opt.coordinate = 'ShoeV2_Coordinate'
opt.roor_dir = './Sketchy'
opt.mode = 'Train'
opt.Train = True
opt.shuffle = True
opt.nThreads = 1
opt.batchsize = 3
dataset_sketchy = CreateDataset_Sketchy(opt, on_Fly=False)
dataloader_sketchy = data.DataLoader(dataset_sketchy, batch_size=opt.batchsize, shuffle=opt.shuffle,
num_workers=int(opt.nThreads))
for i_batch, sanpled_batch in enumerate(dataloader_sketchy):
t0 = time.time()
if i_batch == 0:
print(len(sanpled_batch['sketch_img']))
torchvision.utils.save_image(sanpled_batch['sketch_img'], 'sketch_img.jpg', normalize=True)
torchvision.utils.save_image(sanpled_batch['positive_img'], 'positive_img.jpg', normalize=True)
torchvision.utils.save_image(sanpled_batch['negetive_img'], 'negetive_img.jpg', normalize=True)
print(sanpled_batch['sketch_img'][0].shape)
for i_num in range(len(sanpled_batch['sketch_img'])):
torchvision.utils.save_image(sanpled_batch['sketch_img'][i_num], str(i_num) + 'sketch_img.jpg',
normalize=True)