-
Notifications
You must be signed in to change notification settings - Fork 3
/
data_loader.py
260 lines (191 loc) · 9.01 KB
/
data_loader.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
from torch.utils import data
from torchvision import transforms as T
from PIL import Image
import os
import numpy as np
import pickle
from glob import glob
DEBUG=False
def central_crop(img):
size = min(img.shape[0], img.shape[1])
offset_h = int((img.shape[0] - size) / 2)
offset_w = int((img.shape[1] - size) / 2)
return img[offset_h:offset_h + size, offset_w:offset_w + size]
class MedicalData(data.Dataset):
def __init__(self, image_dir, transform, mode):
self.image_dir = image_dir
self.transform = transform
self.mode = mode
self.datasetA = []
self.datasetB = []
self.preprocess()
if mode == 'train':
self.num_images = len(self.datasetA) + len(self.datasetB)
else:
self.num_images = max(len(self.datasetA), len(self.datasetB))
def preprocess(self):
if self.mode in ['train'] :
pos = glob(os.path.join(self.image_dir, 'train', 'pos', '*png'))
neg = glob(os.path.join(self.image_dir, 'train', 'neg', '*png'))
neg_mixed = glob(os.path.join(self.image_dir, 'train', 'neg_mixed', '*png'))
self.datasetA = pos + neg_mixed
self.datasetB = neg
else:
self.datasetA = glob(os.path.join(self.image_dir, 'test', 'pos', '*png'))
self.datasetB = glob(os.path.join(self.image_dir, 'test', 'neg', '*png'))
print('Finished preprocessing the dataset...')
def __getitem__(self, index):
datasetA = self.datasetA
datasetB = self.datasetB
filenameA = datasetA[index%len(datasetA)]
filenameB = datasetB[index%len(datasetB)]
if self.mode in ['train']:
imageA = Image.open(filenameA).convert("RGB")
imageB = Image.open(filenameB).convert("RGB")
else:
imageA = Image.open(filenameA).convert("RGB")
imageB = Image.open(filenameB).convert("RGB")
imageA = np.array(imageA)
imageB = np.array(imageB)
if DEBUG: print("Original image size: ", imageA.shape, imageB.shape, "min: ", np.min(imageA), np.min(imageB), "max: ", np.max(imageA), np.max(imageB), "type: ", imageA.dtype, imageB.dtype)
imageA = np.pad(imageA, [(5, 5), (5, 5), (0, 0)], mode='constant', constant_values=0)
imageA = central_crop(imageA)
imageB = np.pad(imageB, [(5, 5), (5, 5), (0, 0)], mode='constant', constant_values=0)
imageB = central_crop(imageB)
if DEBUG: print("Processed image size: ", imageA.shape, imageB.shape, "min: ", np.min(imageA), np.min(imageB), "max: ", np.max(imageA), np.max(imageB), "type: ", imageA.dtype, imageB.dtype)
imageA = Image.fromarray(imageA.astype(np.uint8))
imageB = Image.fromarray(imageB.astype(np.uint8))
return self.transform(imageA), self.transform(imageB)
def __len__(self):
"""Return the number of images."""
return self.num_images
class TestValidInductive(data.Dataset):
"""Dataset class for the inductive testing."""
def __init__(self, image_dir, transform, mode):
self.image_dir = image_dir
self.transform = transform
self.mode = mode
self.datasetA = []
self.datasetB = []
self.preprocess()
if "ano" in self.mode:
self.num_images = len(self.datasetA)
elif "hea" in self.mode:
self.num_images = len(self.datasetB)
def preprocess(self):
self.datasetA = glob(os.path.join(self.image_dir, 'test', 'pos', '*png'))
self.datasetB = glob(os.path.join(self.image_dir, 'test', 'neg', '*png'))
print(f'Finished preprocessing the dataset for {self.mode} ...')
def __getitem__(self, index):
"""Return one image and its corresponding attribute label."""
if "ano" in self.mode:
dataset = self.datasetA
else:
dataset = self.datasetB
filename = dataset[index%len(dataset)]
image = Image.open(filename).convert("RGB")
image = np.array(image)
if DEBUG: print("Original image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = np.pad(image, [(5, 5), (5, 5), (0, 0)], mode='constant', constant_values=0)
image = central_crop(image)
if DEBUG: print("Processed image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = Image.fromarray(image.astype(np.uint8))
return filename, self.transform(image)
def __len__(self):
"""Return the number of images."""
return self.num_images
class TestValidTransductive(data.Dataset):
"""Dataset class for the transductive testing."""
def __init__(self, image_dir, transform, mode):
self.image_dir = image_dir
self.transform = transform
self.mode = mode
self.datasetA = []
self.datasetB = []
self.preprocess()
if "ano" in self.mode:
self.num_images = len(self.datasetA)
elif "hea" in self.mode:
self.num_images = len(self.datasetB)
def preprocess(self):
self.datasetA = glob(os.path.join(self.image_dir, 'test', 'pos', '*png'))
self.datasetB = glob(os.path.join(self.image_dir, 'test', 'neg', '*png'))
print(f'Finished preprocessing the dataset for {self.mode} ...')
def __getitem__(self, index):
"""Return one image and its corresponding attribute label."""
if "ano" in self.mode:
dataset = self.datasetA
else:
dataset = self.datasetB
filename = dataset[index%len(dataset)]
image = Image.open(filename).convert("RGB")
image = np.array(image)
if DEBUG: print("Original image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = np.pad(image, [(5, 5), (5, 5), (0, 0)], mode='constant', constant_values=0)
image = central_crop(image)
if DEBUG: print("Processed image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = Image.fromarray(image.astype(np.uint8))
return filename, self.transform(image)
def __len__(self):
"""Return the number of images."""
return self.num_images
class testAUCp(data.Dataset):
"""Dataset class for the AUCp calculation."""
def __init__(self, image_dir, transform, mode):
self.image_dir = image_dir
self.transform = transform
self.mode = mode
self.datasetA = []
self.datasetB = []
self.preprocess()
if "ano" in self.mode:
self.num_images = len(self.datasetA)
elif "hea" in self.mode:
self.num_images = len(self.datasetB)
def preprocess(self):
self.datasetA = glob(os.path.join(self.image_dir, 'test', 'pos', '*png'))
self.datasetB = glob(os.path.join(self.image_dir, 'test', 'neg', '*png'))
print(f'Finished preprocessing the dataset for {self.mode} ...')
def __getitem__(self, index):
"""Return one image and its corresponding attribute label."""
if "ano" in self.mode:
dataset = self.datasetA
else:
dataset = self.datasetB
filename = dataset[index%len(dataset)]
image = Image.open(filename).convert("RGB")
image = np.array(image)
if DEBUG: print("Original image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = np.pad(image, [(5, 5), (5, 5), (0, 0)], mode='constant', constant_values=0)
image = central_crop(image)
if DEBUG: print("Processed image size: ", image.shape, "min: ", np.min(image), "max: ", np.max(image), "type: ", image.dtype)
image = Image.fromarray(image.astype(np.uint8))
return filename, self.transform(image)
def __len__(self):
"""Return the number of images."""
return self.num_images
def get_loader(image_dir, image_size=192, batch_size=1, dataset='MedicalData', mode='train', num_workers=1):
"""Build and return a data loader."""
transform = []
if mode == 'train':
transform.append(T.RandomHorizontalFlip())
transform.append(T.Resize(image_size))
transform.append(T.ToTensor())
transform.append(T.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)))
transform = T.Compose(transform)
if dataset == 'MedicalData':
dataset = MedicalData(image_dir, transform, mode)
elif dataset == 'TestValidInductive':
dataset = TestValidInductive(image_dir, transform, mode)
elif dataset == 'TestValidTransductive':
dataset = TestValidTransductive(image_dir, transform, mode)
elif dataset == 'testAUCp':
dataset = testAUCp(image_dir, transform, mode)
else:
print("Dataset not found!")
exit()
data_loader = data.DataLoader(dataset=dataset,
batch_size=batch_size,
shuffle=(mode=='train'),
num_workers=num_workers)
return data_loader