-
Notifications
You must be signed in to change notification settings - Fork 3
/
dataset_masks.py
268 lines (224 loc) · 8.36 KB
/
dataset_masks.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
import os
import glob
import pickle
import numpy as np
import pandas as pd
from pathlib import Path
from PIL import Image
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as T
def import_small_mask_dataset(
data_path: str = './data/lung_cancer/ds_iter03/',
num_labels: int = 5,
use_split: bool = True,
size: int = 64,
transform=T.PILToTensor(),
batch_size: int = 32,
num_workers: int = 1,
small=True,
):
if not small:
data_path = data_path.replace('ds_iter03','large_masks')
trainset = SmallMaskDataset(
data_path=data_path,
num_labels=num_labels,
use_split = use_split,
size=size,
mode='train',
transform=transform,
)
testset = SmallMaskDataset(
data_path=data_path,
num_labels=num_labels,
use_split = use_split,
size=size,
mode='test',
transform=transform,
)
train_loader = DataLoader(
trainset, batch_size=batch_size, shuffle=True, num_workers=num_workers
)
test_loader = DataLoader(
testset, batch_size=batch_size, shuffle=True, num_workers=num_workers
)
return train_loader, test_loader
class SmallMaskDataset(Dataset):
def __init__(self,
data_path: str = './data/lung_cancer/ds_iter03/',
num_labels: int = 5,
use_split = True,
df_label_path: str = None,
size: int = 64,
mode: str = 'train',
transform=T.PILToTensor()):
self.data_path = data_path
self.num_labels = num_labels
if use_split:
split_idx_path = os.path.join(data_path,'split_indices',f'{mode}_{self.num_labels-1}classes.pickle')
self.split_idx = pd.read_pickle(split_idx_path)
self.use_split = use_split
self.df_label_path = df_label_path
self.uuid2label = self.extract_labels()
self.mask_path = os.path.join(self.data_path,f'segmasks_{self.num_labels}classes')
if not os.path.isdir(self.mask_path):
Path(self.mask_path).mkdir(parents=True, exist_ok=True)
filter_mask(self.data_path,num_labels=self.num_labels)
self.masks,self.labels = self.get_masks_and_labels()
self.size = size
self.transform = transform
def extract_labels(self):
label_path = os.path.join(self.data_path, 'labels')
if os.path.isdir(label_path):
uuid2label = pd.read_pickle(label_path + f'/labels.pickle')
elif self.df_label_path is not None:
uuid2label = get_label_dict(self.df_label_path)
Path(label_path).mkdir(parents=True, exist_ok=True)
pickle_path = f'{label_path}/labels.pickle'
with open(pickle_path, 'wb') as handle:
pickle.dump(uuid2label, handle, protocol=pickle.HIGHEST_PROTOCOL)
else:
self.df_labels_path = f'{self.data_path}/labels.csv'
uuid2label = get_label_dict(self.df_labels_path)
Path(label_path).mkdir(parents=True, exist_ok=True)
pickle_path = f'{label_path}/labels.pickle'
with open(pickle_path, 'wb') as handle:
pickle.dump(uuid2label, handle, protocol=pickle.HIGHEST_PROTOCOL)
return uuid2label
def get_masks_and_labels(self):
labels = list()
masks = list()
if self.use_split:
for key in self.split_idx:
if key == 0:
continue
list_path = self.split_idx[key]
for mask_name in list_path:
uuid = mask_name.split('.tiff')[0]
label_name = self.uuid2label[uuid]
label = self.get_label(label_name)
mask_path = self.mask_path + f'/{mask_name}_mask.png'
masks.append(mask_path)
labels.append(label)
else:
for mask_name in os.listdir(self.mask_path):
mask_path = os.path.join(self.mask_path,mask_name)
uuid = mask_name.split('.tiff')[0]
label_name = self.uuid2label[uuid]
label = self.get_label(label_name)
masks.append(mask_path)
labels.append(label)
return masks, labels
def get_label(self,label_name):
if 'adeno' in label_name.lower():
label = 0
elif 'squamous' in label_name.lower():
label = 1
elif 'plattenepithel' in label_name.lower():
label = 1
else:
assert False, f'{label_name} is not a valid label name.'
return label
def __repr__(self):
rep = f"{type(self).__name__}: ImageFolderDataset[{len(self.images)}]"
for n, x in enumerate(self.images):
rep += f'\nImg: {x}\t'
if n > 10:
rep += '\n...'
break
return rep
def __len__(self):
return len(self.masks)
def __getitem__(self, idx):
# resize mask to (1,self.size,self.size)
# small areas of labels {1,2,3,4} will be overwritten by label 0
# the smaller self.size, the more labels will be overwritten
mask = Image.open(self.masks[idx]).resize((self.size, self.size), Image.NEAREST)
label = self.labels[idx]
if self.transform is not None:
mask = self.transform(mask)
return mask, label
def filter_mask(data_path,num_lables=5):
'''
if self.num_labels 10 we use the following labels:
0 -> 0 Unknown
2 -> 1 Alveole
4 -> 2 Artery
5 -> 3 Artifacts
6 -> 4 Carcinoma
7 -> 5 Cartilage
9 -> 6 Connections
8,14-> 7 Necrosis
18 -> 8 Tumor stroma
other values-> 9 Others
if self.num_labels 5 we use the following labels:
0 -> 0 Unknown
7 -> 1 Carcinoma
9,18-> 2 Necrosis
23 -> 3 Tumor stroma
other values-> 4 Others
'''
folder_path = os.path.join(data_path,'images')
filelist = glob.glob(f'{folder_path}/*mask.png')
save_folder = os.path.join(data_path,f'segmasks_{num_labels}classes')
for mask_path in filelist:
print(f'filter mask at {mask_path}')
mask_name = mask_path.split('/')[-1]
mask = np.array(Image.open(mask_path))
clean_mask = np.zeros_like(mask)
if num_labels==10:
#2->1 Alveole: 2
clean_mask[mask==2]=1
mask[mask==2] = 0
#4->2 Artery: 4
clean_mask[mask == 4] = 2
mask[mask == 4] = 0
#5->3 Artifact: 5
clean_mask[mask == 5] = 3
mask[mask == 5] = 0
#7->4 Carcinoma: 7
clean_mask[mask == 7] = 4
mask[mask == 7] = 0
#8->5 Cartilage: 8
clean_mask[mask == 8] = 5
mask[mask == 8] = 0
#10->6 Connective Tissue: 10
clean_mask[mask == 10] = 6
mask[mask == 10] = 0
#9,18->7 Cells undergoing necrosis: 9 and Necrosis: 18
clean_mask[mask == 9] = 7
mask[mask == 9] = 0
clean_mask[mask == 18] = 7
mask[mask == 18] = 0
#23->8 Tumor stroma: 23
clean_mask[mask == 23] = 8
mask[mask == 23] = 0
#other values->9 Others
clean_mask[mask !=0] = 9
elif num_labels==5:
#7->1 Carcinoma
clean_mask[mask == 7] = 1
mask[mask == 7] = 0
#9,18->2 Necrosis
clean_mask[mask == 9] = 2
mask[mask == 9] = 0
clean_mask[mask == 18] = 2
mask[mask == 18] = 0
#23->3 Tumor stroma
clean_mask[mask == 23] = 3
mask[mask == 23] = 0
#other values->4 Others
clean_mask[mask !=0] = 4
else:
assert False, f'No label specification for {num_labels} classes.'
save_path = f'{save_folder}/{mask_name}'
Image.fromarray(clean_mask).save(save_path)
return
def get_label_dict(df_path):
df = pd.read_csv(df_path)
uuid2label = dict()
for uuid, subtype in zip(df['uuid'], df['subtype']):
new_key = uuid.split('/')[1]
if new_key in uuid2label.keys():
continue
uuid2label[new_key] = subtype
return uuid2label