|
1
|
|
|
""" |
|
2
|
|
|
Download the demo data and sort them into train, val and test in h5 files |
|
3
|
|
|
""" |
|
4
|
|
|
import os |
|
5
|
|
|
import shutil |
|
6
|
|
|
import zipfile |
|
7
|
|
|
|
|
8
|
|
|
import h5py |
|
9
|
|
|
from scipy import ndimage |
|
10
|
|
|
from tensorflow.keras.utils import get_file |
|
11
|
|
|
|
|
12
|
|
|
PROJECT_DIR = "demos/grouped_mask_prostate_longitudinal" |
|
13
|
|
|
os.chdir(PROJECT_DIR) |
|
14
|
|
|
|
|
15
|
|
|
DATA_PATH = "dataset" |
|
16
|
|
|
ZIP_FILE = "data" |
|
17
|
|
|
ORIGIN = "https://github.com/YipengHu/example-data/raw/master/longi-masks/data.zip" |
|
18
|
|
|
|
|
19
|
|
|
if os.path.exists(DATA_PATH): |
|
20
|
|
|
shutil.rmtree(DATA_PATH) |
|
21
|
|
|
os.mkdir(DATA_PATH) |
|
22
|
|
|
|
|
23
|
|
|
zip_file = os.path.join(DATA_PATH, ZIP_FILE + ".zip") |
|
24
|
|
|
get_file(os.path.abspath(zip_file), ORIGIN) |
|
25
|
|
|
with zipfile.ZipFile(zip_file, "r") as zf: |
|
26
|
|
|
zf.extractall(DATA_PATH) |
|
27
|
|
|
os.remove(zip_file) |
|
28
|
|
|
|
|
29
|
|
|
print("\nMask data downloaded: %s." % os.path.abspath(DATA_PATH)) |
|
30
|
|
|
|
|
31
|
|
|
## now read the data and convert to train/val/test |
|
32
|
|
|
ratio_val = 0.1 |
|
33
|
|
|
ratio_test = 0.2 |
|
34
|
|
|
|
|
35
|
|
|
data_filename = os.path.join(DATA_PATH, ZIP_FILE + ".h5") |
|
36
|
|
|
fid_data = h5py.File(data_filename, "r") |
|
37
|
|
|
num_data = len(fid_data) |
|
38
|
|
|
ids_group, ids_ob = [], [] |
|
39
|
|
|
for f in fid_data: |
|
40
|
|
|
ds, ig, io = fid_data[f].name.split("-") |
|
41
|
|
|
if ds == "/group": |
|
42
|
|
|
ids_group.append(int(ig)) |
|
43
|
|
|
ids_ob.append(int(io)) |
|
44
|
|
|
ids_group_unique = list(set(ids_group)) |
|
45
|
|
|
num_group = len(ids_group_unique) |
|
46
|
|
|
num_val = int(num_group * ratio_val) |
|
47
|
|
|
num_test = int(num_group * ratio_test) |
|
48
|
|
|
num_train = num_group - num_val - num_test |
|
49
|
|
|
|
|
50
|
|
|
print("Found %d data in %d groups." % (num_data, num_group)) |
|
51
|
|
|
print( |
|
52
|
|
|
"Dividing into %d-%d-%d for train-val-test (%0.2f-%0.2f-%0.2f)..." |
|
53
|
|
|
% ( |
|
54
|
|
|
num_train, |
|
55
|
|
|
num_val, |
|
56
|
|
|
num_test, |
|
57
|
|
|
1 - ratio_val - ratio_test, |
|
58
|
|
|
ratio_val, |
|
59
|
|
|
ratio_test, |
|
60
|
|
|
) |
|
61
|
|
|
) |
|
62
|
|
|
|
|
63
|
|
|
# write |
|
64
|
|
|
fid_image, fid_label = [], [] |
|
65
|
|
|
folders = [ |
|
66
|
|
|
os.path.join(DATA_PATH, "train"), |
|
67
|
|
|
os.path.join(DATA_PATH, "val"), |
|
68
|
|
|
os.path.join(DATA_PATH, "test"), |
|
69
|
|
|
] |
|
70
|
|
|
for fn in folders: |
|
71
|
|
|
os.mkdir(fn) |
|
72
|
|
|
fid_label.append(h5py.File(os.path.join(fn, "labels.h5"), "w")) |
|
73
|
|
|
fid_image.append(h5py.File(os.path.join(fn, "images.h5"), "w")) |
|
74
|
|
|
|
|
75
|
|
|
for i in range(num_data): |
|
76
|
|
|
dataset_name = "group-%d-%d" % (ids_group[i], ids_ob[i]) |
|
77
|
|
|
pos_group = ids_group_unique.index(ids_group[i]) |
|
78
|
|
|
if pos_group < num_train: # train |
|
79
|
|
|
idf = 0 |
|
80
|
|
|
elif pos_group < (num_train + num_val): # val |
|
81
|
|
|
idf = 1 |
|
82
|
|
|
else: # test |
|
83
|
|
|
idf = 2 |
|
84
|
|
|
data = fid_data[dataset_name] |
|
85
|
|
|
fid_label[idf].create_dataset( |
|
86
|
|
|
dataset_name, shape=data.shape, dtype=data.dtype, data=data |
|
87
|
|
|
) |
|
88
|
|
|
fid_label[idf].flush() |
|
89
|
|
|
image = ndimage.gaussian_filter( |
|
90
|
|
|
data, sigma=3, output="float32" |
|
91
|
|
|
) # smoothing with gaussian |
|
92
|
|
|
fid_image[idf].create_dataset( |
|
93
|
|
|
dataset_name, shape=image.shape, dtype=image.dtype, data=image |
|
94
|
|
|
) |
|
95
|
|
|
fid_image[idf].flush() |
|
96
|
|
|
# print(idf,dataset_name) |
|
97
|
|
|
|
|
98
|
|
|
# close all |
|
99
|
|
|
fid_data.close() |
|
100
|
|
|
for idf in range(len(folders)): |
|
101
|
|
|
fid_label[idf].close() |
|
102
|
|
|
fid_image[idf].close() |
|
103
|
|
|
os.remove(data_filename) |
|
104
|
|
|
|
|
105
|
|
|
print("Done. \n") |
|
106
|
|
|
|
|
107
|
|
|
## now download the pretrained model |
|
108
|
|
|
MODEL_PATH = os.path.join(DATA_PATH, "pretrained") |
|
109
|
|
|
if os.path.exists(MODEL_PATH): |
|
110
|
|
|
shutil.rmtree(MODEL_PATH) |
|
111
|
|
|
os.mkdir(MODEL_PATH) |
|
112
|
|
|
|
|
113
|
|
|
ZIP_PATH = "grouped_mask_prostate_longitudinal_1" |
|
114
|
|
|
ORIGIN = "https://github.com/DeepRegNet/deepreg-model-zoo/raw/master/demo/grouped_mask_prostate_longitudinal/20210110.zip" |
|
115
|
|
|
|
|
116
|
|
|
zip_file = os.path.join(MODEL_PATH, ZIP_PATH + ".zip") |
|
117
|
|
|
get_file(os.path.abspath(zip_file), ORIGIN) |
|
118
|
|
|
with zipfile.ZipFile(zip_file, "r") as zf: |
|
119
|
|
|
zf.extractall(path=MODEL_PATH) |
|
120
|
|
|
os.remove(zip_file) |
|
121
|
|
|
|
|
122
|
|
|
print( |
|
123
|
|
|
"pretrained model is downloaded and unzipped in %s." % os.path.abspath(MODEL_PATH) |
|
124
|
|
|
) |
|
125
|
|
|
|
DeepRegNet /
DeepReg
