-
Notifications
You must be signed in to change notification settings - Fork 0
/
datasets.py
93 lines (72 loc) · 3.25 KB
/
datasets.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
import cv2
import random
import numpy as np
import SimpleITK as sitk
import torch
from torch.utils.data import Dataset
import torchvision.transforms.functional as TF
class BeginDetectorDataset(Dataset):
def __init__(self, df, pos_num, neg_mul, mode='test', size=1000, shape=(512, 512), expand=False):
self.df = df
self.mode= mode
self.size = size
self.shape = shape
self.expand = expand
self.pos_num = pos_num
self.neg_mul = neg_mul
self.neg_num = int(pos_num*neg_mul)
self.batch_size = self.pos_num + self.neg_num
self.norm = torch.nn.InstanceNorm2d(1)
def __len__(self):
return self.size
def __getitem__(self, i):
pos_batch = []
neg_batch = []
while len(pos_batch) + len(neg_batch) < self.batch_size:
row = self.df.loc[random.randint(0, len(self.df)-1)]
image = sitk.GetArrayFromImage(sitk.ReadImage(row.Name)).astype(np.float32)
begin_up = row.Begin
begin_down = row['End1vertebra '] if not isinstance(row['End1vertebra '], str) else 0
pos_indices = list(range(begin_down, begin_up+1))
random.shuffle(pos_indices)
neg_indices = list(range(0, begin_down)) + list(range(begin_up+1, image.shape[0]))
random.shuffle(neg_indices)
while len(pos_batch) < self.pos_num and len(pos_indices)>0:
i = pos_indices.pop()
tmp = cv2.resize(image[i], self.shape, interpolation=cv2.INTER_AREA)
tmp = torch.tensor(tmp).unsqueeze(0)
if self.mode == 'train':
pos_batch.append(TF.rotate(tmp, random.randint(1,270), fill=tmp.min().item()))
else:
pos_batch.append(tmp)
while len(neg_batch) < self.neg_num and len(neg_indices)>0:
i = neg_indices.pop()
tmp = cv2.resize(image[i], self.shape, interpolation=cv2.INTER_AREA)
tmp = torch.tensor(tmp).unsqueeze(0)
neg_batch.append(tmp)
batch = pos_batch + neg_batch
targets = [1]*len(pos_batch) + [0]*len(neg_batch)
for_shuffle = list(zip(batch, targets))
random.shuffle(for_shuffle)
batch, targets = zip(*for_shuffle)
batch = torch.stack(batch)
batch = self.norm(batch)
if self.expand:
batch = batch.expand(batch.shape[0], 3, batch.shape[2], batch.shape[3])
targets = torch.tensor(targets)
return batch, targets
class BeginDetectorDataset_for_infer(Dataset):
def __init__(self, file_path, shape=(512, 512), expand=False):
self.shape = shape
self.expand = expand
self.file_path = file_path
self.norm = torch.nn.InstanceNorm2d(1)
self.image = sitk.GetArrayFromImage(sitk.ReadImage(self.file_path)).astype(np.float32)
def __len__(self):
return self.image.shape[0]
def __getitem__(self, i):
slide = torch.tensor(cv2.resize(self.image[i, ...], self.shape, interpolation=cv2.INTER_AREA)).unsqueeze(0).unsqueeze(0) #
slide = self.norm(slide)
if self.expand:
slide = slide.expand(1, 3, slide.shape[2], slide.shape[3])
return slide, i