-
Notifications
You must be signed in to change notification settings - Fork 1
/
instance_seg_dataset.py
263 lines (225 loc) · 10.4 KB
/
instance_seg_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
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
import logging
import os
import random
import numpy as np
import torch
from PIL import Image
import cv2
import dataloaders.utils as utils
class cell_lab_dataset(torch.utils.data.Dataset):
def __init__(self, root, split, transforms, cache_labels=False, shuffle=False, patch_size=[512, 512],
need_seam_less_clone=False,
seam_less_clone_k_size=(71, 71),
seam_less_blur_sigma=30,
blob_detection=True,
dictionary_mapping={'alive': 1, 'inhib': 2, 'dead': 3}, is_test=False):
self.root = root
self.transforms = transforms
self.split = split
self.blob_detection = blob_detection
self.patch_size = patch_size
self.cache_labels = cache_labels
self.need_seam_less_clone = need_seam_less_clone
self.seam_less_clone_k_size = seam_less_clone_k_size
self.seam_less_blur_sigma = seam_less_blur_sigma
self.dictionary_mapping = dictionary_mapping
self.is_test = is_test
self.bboxes_path_or_cache = []
self.images_path_or_cache = []
self.img_labels_path_or_cache = []
self.images_path = []
for cell_type in self.split:
bboxes_dir_path = os.path.join(self.root, 'weak_labels_reduced_nms')
image_dir_path = os.path.join(root, cell_type)
images_dir = os.listdir(image_dir_path)
img_list = list(
sorted([os.path.join(cell_type, string) for string in images_dir if string.endswith(".jpg")]))
bboxes_path_or_cache = []
images_path_or_cache = []
img_labels_path_or_cache = []
for cell_name in img_list:
bboxes_path = os.path.join(bboxes_dir_path, cell_name.split('.')[-2] + '.txt')
img_path = os.path.join(self.root, cell_name)
if cache_labels:
annotations = np.loadtxt(bboxes_path,
dtype={'names': ('cell_name', 'x_min', 'y_min', 'x_max', 'y_max'),
'formats': ('U25', 'i4', 'i4', 'i4', 'i4')}, delimiter=' ')
boxes = np.dstack(
(annotations['x_min'], annotations['y_min'], annotations['x_max'], annotations['y_max']))
boxes = boxes[0].tolist()
bboxes_path_or_cache.append(boxes)
img = cv2.imread(img_path, cv2.IMREAD_COLOR)
images_path_or_cache.append(img)
labels_mapped = list(map(dictionary_mapping.get, annotations['cell_name'].reshape(-1).tolist()))
img_labels_path_or_cache.append(labels_mapped)
else:
images_path_or_cache.append(img_path)
bboxes_path_or_cache.append(bboxes_path)
img_labels_path_or_cache.append(bboxes_path)
self.images_path_or_cache += images_path_or_cache
self.bboxes_path_or_cache += bboxes_path_or_cache
self.img_labels_path_or_cache += img_labels_path_or_cache
self.images_path += img_list
self.sample_list = list(zip(self.images_path_or_cache, self.bboxes_path_or_cache,self.images_path))
def __getitem__(self, idx):
# load images and masks
img_path = self.sample_list[idx][2]
if self.cache_labels:
# cell_name =
boxes = self.sample_list[idx][1]
img = self.sample_list[idx][0]
# Todo add to the sample list
labels = self.img_labels_path_or_cache[idx]
else:
bboxes_path = self.sample_list[idx][1]
annotations = np.loadtxt(bboxes_path,
dtype={'names': ('cell_name', 'x_min', 'y_min', 'x_max', 'y_max'),
'formats': ('U25', 'i4', 'i4', 'i4', 'i4')}, delimiter=' ')
boxes = np.dstack((annotations['x_min'], annotations['y_min'], annotations['x_max'], annotations['y_max']))
boxes = boxes[0].tolist()
img = cv2.imread(self.sample_list[idx][0], cv2.IMREAD_COLOR)
labels = list(map(self.dictionary_mapping.get, annotations['cell_name'].tolist()))
boxes_post_process = []
labels_post_process = []
img_mask = np.zeros_like(img)
for box, label in zip(boxes, labels):
xmin = int(box[0])
ymin = int(box[1])
xmax = int(box[2])
ymax = int(box[3])
if self.is_test or (xmin < xmax and ymin < ymax and xmin >= 0 and ymin >= 0 and xmax < img.shape[
1] and ymax < img.shape[0]):
boxes_post_process.append(box)
labels_post_process.append(label)
img_mask[ymin:ymax, xmin:xmax, :] = 1
else:
# print(xmin, xmax, ymin, ymax, img.shape)
pass
if self.need_seam_less_clone:
img = utils.seam_less_clone(img, img_mask, ksize=self.seam_less_clone_k_size,
sigma=self.seam_less_blur_sigma)
boxes = boxes_post_process
labels = labels_post_process
if self.blob_detection:
labels = [1] * len(boxes)
target = {}
target['image_size'] = torch.as_tensor([img.shape[0], img.shape[1]], dtype=torch.int64)
if self.transforms is not None:
img_np = np.array(img)
if not img_np.dtype == np.uint8:
logging.info("Error: Image is not of type np.uint8?")
raise
img_np = img_np.astype(np.float32) / 255
result = self.transforms(
image=img_np, bboxes=boxes, category_id=labels)
boxes = result['bboxes']
num_objs = len(boxes)
img = result['image']
# convert everything into a torch.Tensor
if len(boxes) != 0:
boxes = torch.as_tensor(boxes, dtype=torch.float32)
labels = torch.as_tensor(result['category_id'], dtype=torch.int64)
else:
print('image {} with no boxes'.format(img_path))
boxes = torch.empty((0, 4), dtype=torch.float32)
labels = torch.empty((0), dtype=torch.int64)
area = (boxes[:, 3] - boxes[:, 1]) * (boxes[:, 2] - boxes[:, 0])
masks = torch.empty((0,), dtype=torch.uint8)
image_id = torch.tensor([idx])
# suppose all instances are not crowd
iscrowd = torch.zeros((num_objs,), dtype=torch.int64)
target["boxes"] = boxes
target["labels"] = labels
# target["masks"] = masks
target["image_id"] = image_id
target["area"] = area
target["iscrowd"] = iscrowd
return img, target, 0
def __len__(self):
return len(self.sample_list)
class cell_pose_dataset(torch.utils.data.Dataset):
def __init__(self, root, split, transforms):
self.root = root
self.transforms = transforms
self.split = split
# load all image files, sorting them to
# ensure that they are aligned
all = os.listdir(os.path.join(root, split))
self.imgs = list(
sorted([string for string in all if string.endswith("img.png")]))
self.masks = list(
sorted([string for string in all if string.endswith("masks.png")]))
def __getitem__(self, idx):
# load images and masks
img_path = os.path.join(self.root, self.split, self.imgs[idx])
mask_path = os.path.join(self.root, self.split, self.masks[idx])
cell_name = self.imgs[idx]
img = cv2.imread(img_path, cv2.IMREAD_COLOR)
# img = np.stack((img,) * 3, axis=-1)
# note that we haven't converted the mask to RGB,
# because each color corresponds to a different instance
# with 0 being background
mask = cv2.imread(mask_path, cv2.IMREAD_UNCHANGED)
# convert the PIL Image into a numpy array
mask = np.array(mask, np.int16)
# mask = np.array(mask)
target = {}
target['image_size'] = torch.as_tensor([img.shape[0], img.shape[1]], dtype=torch.int64)
if self.transforms is not None:
img_np = np.array(img)
if not img_np.dtype == np.uint8:
logging.info("Error: Image is not of type np.uint8?")
raise
img_np = img_np.astype(np.float32) / 255
result = self.transforms(
image=img_np, mask=mask)
# check images/mask shapes before masks [N, H, W], make mask channel first in tensor
img = result['image']
mask = np.asarray(result['mask'])
# instances are encoded as different colors
obj_ids = np.unique(mask)
# first id is the background, so remove it
obj_ids = obj_ids[1:]
# split the color-encoded mask into a set
# of binary masks
masks = mask == obj_ids[:, None, None]
num_objs = len(obj_ids)
boxes = []
invalid_ids = []
for i in range(num_objs):
pos = np.where(masks[i])
xmin = np.min(pos[1])
xmax = np.max(pos[1])
ymin = np.min(pos[0])
ymax = np.max(pos[0])
# checking degenerated boxes or ugly boxes
if xmin < xmax and ymin < ymax:
boxes.append([xmin, ymin, xmax, ymax])
else:
invalid_ids.append(i)
masks = np.delete(masks, invalid_ids, axis=0)
labels = torch.ones((num_objs - len(invalid_ids),), dtype=torch.int64)
# convert everything into a torch.Tensor
if len(boxes) != 0:
boxes = torch.as_tensor(boxes, dtype=torch.float32)
else:
print('image {} with no boxes'.format(img_path))
boxes = torch.empty((0, 4), dtype=torch.float32)
# there is only one class
masks = torch.as_tensor(masks, dtype=torch.uint8)
image_id = torch.tensor([idx])
# suppose all instances are not crowd
iscrowd = torch.zeros((num_objs,), dtype=torch.int64)
if torch.numel(boxes) != 0:
area = (boxes[:, 3] - boxes[:, 1]) * (boxes[:, 2] - boxes[:, 0])
else:
area = boxes
target["boxes"] = boxes
target["labels"] = labels
target["masks"] = masks
target["image_id"] = image_id
target["area"] = area
target["iscrowd"] = iscrowd
return img, target, cell_name
def __len__(self):
return len(self.imgs)