-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathutils.py
316 lines (257 loc) · 9.96 KB
/
utils.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
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
# Written by Tiankai Yang
import os
import time
import numpy as np
from PIL import Image
import torch
from torchvision import transforms
import cv2
from sklearn.model_selection import train_test_split
from kornia.color import rgb_to_lab
import gc
from config import DefaultConfig
config = DefaultConfig()
device = DefaultConfig.device
# torch.set_default_dtype(torch.float32)
# Modified from https://github.com/zohrehazizi/torch_SSL by Tiankai Yang
class TimerClass():
def __init__(self, report_timing=True):
self.events = ['start']
self.times = [time.time()]
self.report_timing = report_timing
def register(self, eventname):
self.events.append(eventname)
self.times.append(time.time())
print(f"{eventname}: {self.times[-1]-self.times[-2]}")
def print(self):
if self.report_timing:
totalTime = self.times[-1]-self.times[0]
# save to file in config.temp_dir
with open(os.path.join(DefaultConfig.temp_dir, "timing.txt"), "w") as f:
f.write(f"total time: {totalTime}\n")
for i in range(1, len(self.events)):
event = self.events[i]
time = self.times[i]
time_prev = self.times[i-1]
f.write(
f"{event}: {time-time_prev}, percentage: {(time-time_prev)/(totalTime)*100}\n")
def convert_to_torch(X):
if type(X)!=torch.Tensor:
return torch.tensor(X, dtype=torch.float32, device=device)
if X.device != device:
return X.to(device)
return X
def convert_to_numpy(x):
if type(x)==torch.Tensor:
return x.detach().cpu().numpy()
else:
return x
def upsample_x2_lanczos(input_set):
output_set = np.zeros((input_set.shape[0], input_set.shape[1], input_set.shape[2]*2, input_set.shape[3]*2))
for i in range(input_set.shape[0]):
for j in range(input_set.shape[1]):
output_set[i,j,:,:] = cv2.resize(src=input_set[i,j,:,:], dsize=None,
fx=2, fy=2, interpolation=cv2.INTER_LANCZOS4)
# output_set[i,j,:,:] = cv2.resize(src=input_set[i,j,:,:], dsize=None,
# fx=2, fy=2, interpolation=cv2.INTER_LINEAR)
return output_set
def RGB2YUV(image_set):
# image_set: Nx3xHxW
Y_trans_matrix = torch.tensor([0.299, 0.587, 0.114])
U_trans_matrix = torch.tensor([-0.14713, -0.28886, 0.436])
V_trans_matrix = torch.tensor([0.615, -0.51499, -0.10001])
h, w = image_set.shape[2], image_set.shape[3]
image_set = image_set.view(image_set.shape[0], 3, -1)
Y_set = torch.matmul(Y_trans_matrix, image_set).view(image_set.shape[0], 1, h, w)
U_set = torch.matmul(U_trans_matrix, image_set).view(image_set.shape[0], 1, h, w)
V_set = torch.matmul(V_trans_matrix, image_set).view(image_set.shape[0], 1, h, w)
del image_set
del Y_trans_matrix
del U_trans_matrix
del V_trans_matrix
gc.collect()
# return Y_set, U_set, V_set
return torch.cat([Y_set, U_set, V_set], dim=1)
def RGB2LAB(image_set):
# image_set: Nx3xHxW
lab_set = rgb_to_lab(image_set)
return lab_set
def load_casia_v2_lab(index=None):
name_list = os.listdir(config.casia_v2_image_dir)
# only keep .png files and remove extension
name_list = [name for name in name_list if name.endswith(".png")]
print("number of images: ", len(name_list))
name_list.sort()
# load images, surfaces and edges
image_set = []
surface_set = []
edge_set = []
if index is not None:
# if index is a list
if type(index) == list:
name_list = [name_list[i] for i in index]
# if index is a single number
else:
name_list = [name_list[index]]
print(name_list)
for i in range(len(name_list)):
image = Image.open(os.path.join(
config.casia_v2_image_dir, name_list[i]))
surface = Image.open(os.path.join(
config.casia_v2_surface_dir, name_list[i]))
# edge = Image.open(os.path.join(config.casia_v2_edge_dir, name_list[i]))
# [0, 1]
image = transforms.ToTensor()(image)
surface = transforms.ToTensor()(surface)
# edge = transforms.ToTensor()(edge)
###!!!!!!! handle edge labels
image_set.append(image)
surface_set.append(surface)
if index is not None:
# to tensor
image_set = torch.stack(image_set)
surface_set = torch.stack(surface_set)
# edge_set = torch.stack(edge_set)
# # !turn to lab
# lab_set = RGB2LAB(image_set)
# del image_set
# gc.collect()
# return lab_set, surface_set
# !turn to yuv
image_set = image_set * 255
yuv_set = RGB2YUV(image_set)
del image_set
gc.collect()
return yuv_set, surface_set
# return image_set, surface_set
test_image_set, image_set, test_surface_set, surface_set = \
train_test_split(image_set, surface_set, test_size=0.8, shuffle=False)
# _, image_set, _, surface_set = \
# train_test_split(image_set, surface_set, test_size=0.75, shuffle=True)
gc.collect()
# to tensor
image_set = torch.stack(image_set)
surface_set = torch.stack(surface_set)
test_image_set = torch.stack(test_image_set)
test_surface_set = torch.stack(test_surface_set)
# # !turn to lab
# lab_set = RGB2LAB(image_set)
# test_lab_set = RGB2LAB(test_image_set)
# del image_set
# del test_image_set
# gc.collect()
# return lab_set, surface_set, test_lab_set, test_surface_set
# !turn to yuv
image_set = image_set * 255
test_image_set = test_image_set * 255
yuv_set = RGB2YUV(image_set)
test_yuv_set = RGB2YUV(test_image_set)
del image_set
del test_image_set
gc.collect()
return yuv_set, surface_set, test_yuv_set, test_surface_set
# return image_set, surface_set, test_image_set, test_surface_set
def load_casia_v2_yuv(index=None):
name_list = os.listdir(config.casia_v2_image_dir)
# only keep .png files and remove extension
name_list = [name for name in name_list if name.endswith(".png")]
print("number of images: ", len(name_list))
name_list.sort()
# load images, surfaces and edges
image_set = []
surface_set = []
edge_set = []
if index is not None:
# if index is a list
if type(index) == list:
name_list = [name_list[i] for i in index]
# if index is a single number
else:
name_list = [name_list[index]]
print(name_list)
# name_list = name_list[:100]
for i in range(len(name_list)):
image = Image.open(os.path.join(
config.casia_v2_image_dir, name_list[i]))
surface = Image.open(os.path.join(
config.casia_v2_surface_dir, name_list[i]))
# edge = Image.open(os.path.join(config.casia_v2_edge_dir, name_list[i]))
# [0, 255]
image = transforms.ToTensor()(image)
image = image * 255
# [0, 1]
surface = transforms.ToTensor()(surface)
# edge = transforms.ToTensor()(edge)
###!!!!!!! handle edge labels
image_set.append(image)
surface_set.append(surface)
if index is not None:
# to tensor
image_set = torch.stack(image_set)
surface_set = torch.stack(surface_set)
y_set, u_set, v_set = RGB2YUV(image_set)
return y_set, u_set, v_set, surface_set
test_images_set, image_set, test_surfaces_set, surface_set = \
train_test_split(image_set, surface_set, test_size=0.8, shuffle=False)
# _, image_set, _, surface_set = \
# train_test_split(image_set, surface_set, test_size=0.5, shuffle=True)
# test_name_list, name_list, _, _ = \
# train_test_split(name_list, name_list, test_size=0.8, shuffle=False)
# print(test_name_list)
# print(len(test_name_list))
# print(name_list[0])
# print(test_name_list[0])
# print(name_list[-1])
# print(test_name_list[-1])
# print("number of training images: ", len(image_set))
# print("number of testing images: ", len(test_images_set))
gc.collect()
# to tensor
image_set = torch.stack(image_set)
surface_set = torch.stack(surface_set)
test_image_set = torch.stack(test_images_set)
test_surface_set = torch.stack(test_surfaces_set)
# to YUV
y_set, u_set, v_set = RGB2YUV(image_set)
test_y_set, test_u_set, test_v_set = RGB2YUV(test_image_set)
return y_set, u_set, v_set, surface_set, \
test_y_set, test_u_set, test_v_set, test_surface_set
def load_columbia_yuv(index=None):
name_list = os.listdir(config.columbia_image_dir)
name_list = [name for name in name_list if name.endswith(".png")]
print("number of images: ", len(name_list))
name_list.sort()
# load images, surfaces and edges
image_set = []
surface_set = []
edge_set = []
if index is not None:
# if index is a list
if type(index) == list:
name_list = [name_list[i] for i in index]
# if index is a single number
else:
name_list = [name_list[index]]
for i in range(len(name_list)):
image = Image.open(os.path.join(
config.columbia_image_dir, name_list[i]))
surface = Image.open(os.path.join(
config.columbia_surface_dir, name_list[i]))
# edge = Image.open(os.path.join(config.casia_v2_edge_dir, name_list[i]))
# [0, 255]
image = transforms.ToTensor()(image)
image = image * 255
# [0, 1]
surface = transforms.ToTensor()(surface)
# edge = transforms.ToTensor()(edge)
###!!!!!!! handle edge labels
image_set.append(image)
surface_set.append(surface)
# to tensor
image_set = torch.stack(image_set)
surface_set = torch.stack(surface_set)
# to YUV
yuv_set = RGB2YUV(image_set)
if index is not None:
print(name_list[0])
return yuv_set, surface_set