-
Notifications
You must be signed in to change notification settings - Fork 0
/
Dataloader.py
151 lines (135 loc) · 4.72 KB
/
Dataloader.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
from torch.utils.data import Dataset
import cv2
import pickle
import gzip
from scipy import ndimage
import os
import numpy as np
import random
import torch
from scipy.ndimage.interpolation import zoom
equivalent_classes = {
# Acevedo-20 dataset
'BA': 'basophil',
'EO': 'eosinophil',
'ERB': 'erythroblast',
'IG': "unknown", # immature granulocytes,
'PMY': 'promyelocyte', # immature granulocytes,
'MY': 'myelocyte', # immature granulocytes,
'MMY': 'metamyelocyte', # immature granulocytes,
'LY': 'lymphocyte_typical',
'MO': 'monocyte',
'NEUTROPHIL': "unknown",
'BNE': 'neutrophil_banded',
'SNE': 'neutrophil_segmented',
'PLATELET': "unknown",
# Matek-19 dataset
'BAS': 'basophil',
'EBO': 'erythroblast',
'EOS': 'eosinophil',
'KSC': 'smudge_cell',
'LYA': 'lymphocyte_atypical',
'LYT': 'lymphocyte_typical',
'MMZ': 'metamyelocyte',
'MOB': 'monocyte', # monoblast
'MON': 'monocyte',
'MYB': 'myelocyte',
'MYO': 'myeloblast',
'NGB': 'neutrophil_banded',
'NGS': 'neutrophil_segmented',
'PMB': "unknown",
'PMO': 'promyelocyte',
}
label_map = {
'basophil': 0,
'eosinophil': 1,
'erythroblast': 2,
'myeloblast': 3,
'promyelocyte': 4,
'myelocyte': 5,
'metamyelocyte': 6,
'neutrophil_banded': 7,
'neutrophil_segmented': 8,
'monocyte': 9,
'lymphocyte_typical': 10,
'lymphocyte_atypical': 11,
'smudge_cell': 12,
}
class Train_DataLoader(Dataset):
def __init__(self, datasets_dir):
self.datasets_names = ['Matek','Acevedo']
self.datasets_dir = datasets_dir
dataset_files = np.unique([x for x in os.listdir(self.datasets_dir) if x.split("-")[0] in self.datasets_names and x.split('-')[-1][-6:]=='dat.gz'])
samples = {}
remove_key=[]
for file in dataset_files:
print('loading ',file, '... ', end='')
with gzip.open(os.path.join(self.datasets_dir, file), 'rb') as f:
data = pickle.load(f)
for d in data:
data[d]['dataset'] = file.split('-')[0]
data[d]['label'] = d.split('_')[0]
data[d]['masks'] = self.masks_processing(data[d]['masks'])
samples = {**samples, **data}
print('[done]')
keys = list(samples.keys())
for s in keys:
if equivalent_classes[samples[s]["label"]] == "unknown" or s in remove_key:
samples.pop(s, None)
self.samples=samples
self.keys=list(samples.keys())
def __len__(self):
return len(self.samples)
def __getitem__(self, index):
key = self.keys[index]
label = label_map[equivalent_classes[self.samples[key]['label']]]
img = self.samples[key]['image'].copy()
dataset=self.datasets_names.index(self.samples[key]['dataset'])
masks=self.samples[key]['masks'].copy()
feature=self.samples[key]['feature']
img = np.rollaxis(img, 2, 0)
cropped_img = img.copy()
cropped_img[~np.stack([masks]*3, axis=0).reshape(3, 224, 224)]=128
cropped_img=self.get_box_cropped(cropped_img,masks)
cropped_img = zoom(cropped_img, (1, 128 / cropped_img.shape[1], 128 / cropped_img.shape[2]), order=3)
cropped_img = torch.from_numpy(cropped_img.astype(np.float32)/255)
feature = torch.from_numpy(feature.astype(np.float32))
return feature, cropped_img, label, dataset
def masks_processing(self,masks):
labeled_arr, num_features = ndimage.label(masks)
sizes = ndimage.sum(masks, labeled_arr, range(num_features+1))
max_label = np.argmax(sizes[1:]) + 1
output = np.zeros_like(masks,dtype=bool)
output[labeled_arr == max_label] = True
return output
def get_box_cropped(self,image,masks):
img=image.copy()
h,w = masks.shape
y1=-1
y2=-1
x1=-1
x2=-1
for i in range(h):
if True in masks[i]:
y1=i
break
for i in range(h-1,-1,-1):
if True in masks[i]:
y2=i
break
for i in range(w):
if True in masks[:,i]:
x1=i
break
for i in range(w-1,-1,-1):
if True in masks[:,i]:
x2=i
break
img=img[:, y1:y2, x1:x2]
f=np.full((3, 10, img.shape[2]),128,dtype=np.uint8)
img=np.concatenate([f,img,f],axis=1)
f=np.full((3, img.shape[1], 10),128,dtype=np.uint8)
img=np.concatenate([f,img,f],axis=2)
return img
if __name__=='__main__':
print(Train_DataLoader('/home/anoke/data/Dataset/'))