-
Notifications
You must be signed in to change notification settings - Fork 12
/
Segmentation_BatchDataset.py
154 lines (122 loc) · 5.48 KB
/
Segmentation_BatchDataset.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
"""
Code ideas from https://github.com/Newmu/dcgan and tensorflow mnist dataset reader
"""
import numpy as np
import scipy.misc as misc
import os
import glob
from random import shuffle, randint
class seg_dataset_reader:
path = ""
class_mappings = ""
files = []
images = []
annotations = []
test_images = []
test_annotations = []
batch_offset = 0
epochs_completed = 0
def __init__(self, deepscores_path, max_pages=40, crop=True, crop_size=[1000,1000], test_size=20):
"""
Initialize a file reader for the DeepScores classification data
:param records_list: path to the dataset
sample record: {'image': f, 'annotation': annotation_file, 'filename': filename}
"""
print("Initializing DeepScores Classification Batch Dataset Reader...")
self.path = deepscores_path
self.max_pages = max_pages
self.crop = crop
self.crop_size = crop_size
self.test_size = test_size
images_list = []
images_glob = os.path.join(self.path, "images_png", '*.' + 'png')
images_list.extend(glob.glob(images_glob))
#shuffle image list
shuffle(images_list)
if max_pages is None:
max_pages = len(images_list)
import sys
sys.exit(1)
if max_pages > len(images_list):
print("Not enough data, only " + str(len(images_list)) + " available")
if test_size >= max_pages:
print("Test set too big ("+str(test_size)+"), max_pages is: "+str(max_pages))
import sys
sys.exit(1)
print("Splitting dataset, train: "+str(max_pages-test_size)+" images, test: "+str(test_size)+ " images")
test_image_list = images_list[0:test_size]
train_image_list = images_list[test_size:max_pages]
# test_annotation_list = [image_file.replace("/images_png/", "/pix_annotations_png/") for image_file in test_image_list]
# train_annotation_list = [image_file.replace("/images_png/", "/pix_annotations_png/") for image_file in train_image_list]
self._read_images(test_image_list,train_image_list)
def _read_images(self,test_image_list,train_image_list):
dat_train = [self._transform(filename) for filename in train_image_list]
for dat in dat_train:
self.images.append(dat[0])
self.annotations.append(dat[1])
self.images = np.array(self.images)
self.images = np.expand_dims(self.images, -1)
self.annotations = np.array(self.annotations)
self.annotations = np.expand_dims(self.annotations, -1)
print("Training set done")
dat_test = [self._transform(filename) for filename in test_image_list]
for dat in dat_test:
self.test_images.append(dat[0])
self.test_annotations.append(dat[1])
self.test_images = np.array(self.test_images)
self.test_images = np.expand_dims(self.test_images, -1)
self.test_annotations = np.array(self.test_annotations)
self.test_annotations = np.expand_dims(self.test_annotations, -1)
print("Test set done")
def _transform(self, filename):
image = misc.imread(filename)
annotation = misc.imread(filename.replace("/images_png/", "/pix_annotations_png/"))
print("im working!" + str(randint(0,10)))
if not image.shape[0:2] == annotation.shape[0:2]:
print("input and annotation have different sizes!")
import sys
import pdb
pdb.set_trace()
sys.exit(1)
if image.shape[-1] != 1:
# take mean over color channels, image BW anyways --> fix in dataset creation
image = np.mean(image, -1)
if self.crop:
coord_0 = randint(0, (image.shape[0] - self.crop_size[0]))
coord_1 = randint(0, (image.shape[1] - self.crop_size[1]))
image = image[coord_0:(coord_0+self.crop_size[0]),coord_1:(coord_1+self.crop_size[1])]
annotation = annotation[coord_0:(coord_0 + self.crop_size[0]), coord_1:(coord_1 + self.crop_size[1])]
return [image, annotation]
# from PIL import Image
# im = Image.fromarray(image)
# im.show()
# im = Image.fromarray(annotation)
# im.show()
def get_records(self):
return self.images, self.annotations
def reset_batch_offset(self, offset=0):
self.batch_offset = offset
def get_test_records(self):
return self.test_images, self.test_annotations
def next_batch(self, batch_size):
start = self.batch_offset
self.batch_offset += batch_size
if self.batch_offset > self.images.shape[0]:
# Finished epoch
self.epochs_completed += 1
print("****************** Epochs completed: " + str(self.epochs_completed) + "******************")
# Shuffle the data
perm = np.arange(self.images.shape[0])
np.random.shuffle(perm)
self.images = self.images[perm]
self.annotations = self.annotations[perm]
# Start next epoch
start = 0
self.batch_offset = batch_size
end = self.batch_offset
return self.images[start:end], self.annotations[start:end]
def get_random_batch(self, batch_size):
indexes = np.random.randint(0, self.images.shape[0], size=[batch_size]).tolist()
return self.images[indexes], self.annotations[indexes]
if __name__ == "__main__":
data_reader = seg_dataset_reader("/Users/tugg/datasets/DeepScores")