-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtorch_dataset.py
162 lines (102 loc) · 4.19 KB
/
torch_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
import random
import torch
from PIL import Image
import numpy as np
import cv2
import albumentations as A
import albumentations.pytorch.transforms as APT
class BreastCancerDataSet_Basic(torch.utils.data.Dataset):
def __init__(self, df, path, TARGET, transforms=None, use_patient_id_for_path = True):
super().__init__()
self.df = df
self.path = path
self.transforms = transforms
self.TARGET = TARGET
self.use_patient_id_for_path = use_patient_id_for_path
def load_image(self, i):
if self.use_patient_id_for_path:
pth = f'{self.path}/{self.df.iloc[i].patient_id}/{self.df.iloc[i].image_id}.png'
else:
pth = f'{self.path}/{self.df.iloc[i].image_id}.png'
try:
img = Image.open(pth) #.convert('RGB')
except Exception as ex:
print(pth, ex)
return None
return img
def get_labels(self, i):
label_cancer = torch.as_tensor(self.df.iloc[i][self.TARGET]).float()
return label_cancer
def __getitem__(self, i):
img = self.load_image(i)
if self.transforms is not None:
img = self.transforms(img)
if self.TARGET not in self.df.columns:
return img
return img, self.get_labels(i)
def __len__(self):
return len(self.df)
class BreastCancerDataSet_Mixup(torch.utils.data.Dataset):
def __init__(self, df, path, TARGET, transforms=None):
super().__init__()
self.df = df
self.path = path
self.transforms = transforms
self.TARGET = TARGET
def load_image(self, i):
pth = f'{self.path}/{self.df.iloc[i].patient_id}/{self.df.iloc[i].image_id}.png'
try:
img = Image.open(pth) #.convert('RGB')
except Exception as ex:
print(pth, ex)
return None
return img
def get_labels(self, i):
label_cancer = torch.as_tensor(self.df.iloc[i][self.TARGET]).float()
return label_cancer
def __getitem__(self, i):
i_mixup = np.random.randint(len(self.df))
img = self.load_image(i)
img_mixup = self.load_image(i_mixup)
if self.transforms is not None:
img = self.transforms(img)
img_mixup = self.transforms(img_mixup)
if self.TARGET not in self.df.columns:
return img, img_mixup
return img, img_mixup, self.get_labels(i)
def __len__(self):
return len(self.df)
class BreastCancerDataSet_MultiLabel(torch.utils.data.Dataset):
def __init__(self, df, path, transforms=None, use_patient_id_for_path = True):
super().__init__()
self.df = df
self.path = path
self.transforms = transforms
self.use_patient_id_for_path = use_patient_id_for_path
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
self.one_hot_subtype = torch.as_tensor(le.fit_transform(df.subtype))
self.one_hot_abnormality = torch.as_tensor(le.fit_transform(df.abnormality))
def load_image(self, i):
if self.use_patient_id_for_path:
pth = f'{self.path}/{self.df.iloc[i].patient_id}/{self.df.iloc[i].image_id}.png'
else:
pth = f'{self.path}/{self.df.iloc[i].image_id}.png'
try:
img = Image.open(pth) #.convert('RGB')
except Exception as ex:
print(pth, ex)
return None
return img
def get_labels(self, i):
label_cancer = torch.as_tensor(self.df.iloc[i].y).float()
label_subtype = self.one_hot_subtype[i]
label_abnormality = self.one_hot_abnormality[i]
return label_cancer, label_subtype, label_abnormality
def __getitem__(self, i):
img = self.load_image(i)
if self.transforms is not None:
img = self.transforms(img)
return img, self.get_labels(i)
def __len__(self):
return len(self.df)