-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathdataset.py
More file actions
207 lines (168 loc) · 8.71 KB
/
dataset.py
File metadata and controls
207 lines (168 loc) · 8.71 KB
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
import os
import pandas as pd
import torch
import numpy as np
from torchvision import transforms
from PIL import Image
from skimage.transform import resize
from sklearn.preprocessing import LabelEncoder
class PDD_Img_Dataset(torch.utils.data.Dataset):
def __init__(self, image_path, embedding_path, CSV_path):
self.image_path = image_path
self.embedding_path = embedding_path
self.CSV_file = pd.read_csv(CSV_path)
self.has_source = 'Source' in self.CSV_file.columns # train single dataset = no Source; train all datasets = Source
def encode_labels(self):
column_name = 'Treatment' if 'Treatment' in self.CSV_file.columns else 'pert_name' if 'pert_name' in self.CSV_file.columns else 'Compound'
if column_name is None:
raise ValueError("CSV file must contain 'Treatment', 'pert_name', or 'Compound' column")
self.CSV_file['encoded_labels'] = self.label_encoder.fit_transform(self.CSV_file[column_name])
def load_image(self, img_name, source=None):
if source:
# img_path = os.path.join(self.image_path, source, "original", 'images', img_name)
img_path = os.path.join(self.image_path, source, 'images', img_name.replace('tif', 'png'))
else:
img_path = os.path.join(self.image_path, img_name.replace('tif', 'png'))
# img_path = os.path.join(self.image_path, "original", 'images', img_name)
return Image.open(img_path)
def __getitem__(self, idx):
item = {}
channels = ['DNA', 'ER', 'RNA', 'AGP', 'Mito']
if self.has_source:
source = self.CSV_file.loc[idx, 'Source']
images_list = [self.load_image(self.CSV_file.loc[idx, channel], source) for channel in channels]
else:
images_list = [self.load_image(self.CSV_file.loc[idx, channel]) for channel in channels]
images = np.stack(images_list, axis=0)
resized_image = resize(images, (5, 448, 448), anti_aliasing=True)
preprocess = transforms.Compose([transforms.ToTensor()])
resized_image_tensor = preprocess(resized_image.transpose(1, 2, 0))
item['image'] = resized_image_tensor.float()
return item
def __len__(self):
return self.CSV_file.shape[0]
class PDDDataset(torch.utils.data.Dataset):
def __init__(self, image_path, embedding_path, CSV_path):
self.image_path = image_path
self.embedding_path = embedding_path
self.CSV_file = pd.read_csv(CSV_path)
self.label_encoder = LabelEncoder()
self.encode_labels()
self.has_source = 'Source' in self.CSV_file.columns # train single dataset = no Source; train all datasets = Source
def encode_labels(self):
column_name = 'Treatment' if 'Treatment' in self.CSV_file.columns else 'pert_name' if 'pert_name' in self.CSV_file.columns else 'Compound'
if column_name is None:
raise ValueError("CSV file must contain 'Treatment', 'pert_name', or 'Compound' column")
self.CSV_file['encoded_labels'] = self.label_encoder.fit_transform(self.CSV_file[column_name])
def load_image(self, img_name, source=None):
if source:
# img_path = os.path.join(self.image_path, source, "original", 'images', img_name)
img_path = os.path.join(self.image_path, source, 'images', img_name.replace('tif', 'png'))
else:
img_path = os.path.join(self.image_path, img_name.replace('tif', 'png'))
# img_path = os.path.join(self.image_path, "original", 'images', img_name)
return Image.open(img_path)
def __getitem__(self, idx):
item = {}
channels = ['DNA', 'ER', 'RNA', 'AGP', 'Mito']
if self.has_source:
source = self.CSV_file.loc[idx, 'Source']
images_list = [self.load_image(self.CSV_file.loc[idx, channel], source) for channel in channels]
else:
images_list = [self.load_image(self.CSV_file.loc[idx, channel]) for channel in channels]
images = np.stack(images_list, axis=0)
resized_image = resize(images, (5, 448, 448), anti_aliasing=True)
preprocess = transforms.Compose([transforms.ToTensor()])
resized_image_tensor = preprocess(resized_image.transpose(1, 2, 0))
if self.has_source:
embedding_path = os.path.join(
self.embedding_path,
self.CSV_file.loc[idx, 'Source'].upper(),
str(self.CSV_file.loc[idx, 'Metadata_Plate']),
str(self.CSV_file.loc[idx, 'Metadata_Well']),
str(self.CSV_file.loc[idx, 'Metadata_Site']),
'PhenoProfiler_embeddings.npy'
)
else:
embedding_path = os.path.join(
self.embedding_path,
str(self.CSV_file.loc[idx, 'Metadata_Plate']),
str(self.CSV_file.loc[idx, 'Metadata_Well']),
str(self.CSV_file.loc[idx, 'Metadata_Site']),
'PhenoProfiler_embeddings.npy'
)
# print(embedding_path)
if embedding_path.endswith('.npz'):
with np.load(embedding_path) as data:
features = data["features"]
embedding = np.median(features[~np.isnan(features).any(axis=1)], axis=0)
else:
features = np.load(embedding_path)
embedding = np.median(features[~np.isnan(features).any(axis=1)], axis=0)
# print(embedding.shape)
item['image'] = resized_image_tensor.float()
item['embedding'] = torch.tensor(embedding).float()
item['class'] = torch.tensor(self.CSV_file.loc[idx, 'encoded_labels']).long()
return item
def __len__(self):
return self.CSV_file.shape[0]
class NoendDataset(torch.utils.data.Dataset):
'''
noendDataset(image_path = "/data/boom/cpg0019/broad/",
embedding_path = "/data/boom/cpg0019/broad/workspace_dl/embeddings/105281_zenodo7114558/",
# CSV_path = "/data/boom/cpg0019/broad/workspace_dl/metadata/sc-metadata-fil.csv")
'''
def __init__(self, image_path, embedding_path, CSV_path):
self.image_path = image_path
self.embedding_path = embedding_path
self.CSV_file = pd.read_csv(CSV_path)
self.label_encoder = LabelEncoder()
self.encode_labels()
def encode_labels(self):
if 'Treatment' in self.CSV_file.columns:
column_name = 'Treatment'
elif 'pert_name' in self.CSV_file.columns:
column_name = 'pert_name'
else:
raise ValueError("CSV file must contain either 'Treatment' or 'pert_name' column")
self.CSV_file['encoded_labels'] = self.label_encoder.fit_transform(self.CSV_file[column_name])
def __getitem__(self, idx):
item = {}
# 获取图像位置，然后读取，然后按照坐标拆分，得到5张图象，然后叠加
img_path = self.image_path + str(self.CSV_file.loc[idx, 'Image_Name'])[6:]
All_img = Image.open(img_path)
# print(img_path, All_img.size)
# 分割大图像为6个子图像，每个子图像的尺寸为（160，160）
sub_images = []
for i in range(6):
left = i * 160
upper = 0
right = left + 160
lower = upper + 160
sub_image = All_img.crop((left, upper, right, lower))
sub_images.append(sub_image)
# 按照通道叠加前面5张子图像在一起
combined_image = np.stack(sub_images[:5], axis=0)
# print(combined_image.shape)
resized_image = resize(combined_image, (5, 224, 224), anti_aliasing=True)
# embedding
# /data/boom/cpg0019/broad/workspace_dl/embeddings/105281_zenodo7114558/BBBC022/20585/A01/1
path = os.path.dirname(self.CSV_file.loc[idx, 'Image_Name'][22:])
# embedding_path = os.path.join(self.embedding_path, str(path), 'embedding.npz')
# print("embedding_path:", self.embedding_path, path, embedding_path)
# with open(embedding_path, "rb") as data:
# info = np.load(data)
# cells = np.array(np.copy(info["features"]))
# embedding = cells[~np.isnan(cells).any(axis=1)]
# # embedding = np.median(embedding, axis=0)
# print(idx, embedding.shape)
# embedding = embedding[idx]
# print(idx, embedding.shape)
item['image'] = torch.tensor(resized_image).float() # torch.Size([5, 448, 448])
# item['embedding'] = torch.tensor(embedding).float() # torch.Size([672])
# print(item['embedding'].shape, item['image'].shape)
encoded_labels = self.CSV_file.loc[idx, 'encoded_labels']
item['class'] = torch.tensor(encoded_labels).long()
return item
def __len__(self):
return self.CSV_file.shape[0]