-
Notifications
You must be signed in to change notification settings - Fork 2
/
make_common_dataset.py
205 lines (180 loc) · 8.36 KB
/
make_common_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
import glob
import os
import pandas as pd
import random
from scipy import sparse
from scipy.sparse import save_npz
import shutil
# read the data from the given subjects
subjects = 'all'
signal_type = 'grad'
n_parcels_max = 3
no_parcels = 450
data_dir_base = 'data/train/'
# assign a number from which first subject_id will start
if 'train' in data_dir_base:
subject_id_start = 1
subdir = 'train'
elif 'test' in data_dir_base:
subject_id_start = 6
subdir = 'test'
divide_to = 2 # if 1 data will not be split. Otherwise it might be 2 or 4.
# the data will be equally split to private and public (if set 2) or
# to private/public/test/train (if set to 4)
data_dir_save = 'data/ramp_challenge'
assert divide_to == 1 or divide_to == 2 or divide_to == 4
# combine them in one datafile with each sample: data from electrodes,
# subject_name, signal_type
data_type = '_' + str(no_parcels) + '_' + str(n_parcels_max)
data_dir_all = os.path.join(data_dir_base,
f'data_{signal_type}_all{data_type}')
data_dir = os.path.join(data_dir_base, f'data_{signal_type}_*{data_type}')
if data_dir_save is None:
data_dir_save = data_dir_all
data_dirs = sorted(glob.glob(data_dir))
# if the all_data directory is still in the list of dirs, remove it
if data_dir_all in data_dirs:
data_dirs.remove(data_dir_all)
assert data_dir_all not in data_dirs
random.seed(42)
random.shuffle(data_dirs)
def save_subject_coords(subject_id, subjects, data_type,
csv_file='data/subjects.csv'):
subjects_name = []
for subject in subjects:
subject_data = subject.split('_')
subjects_name.append(subject_data[-3])
subj_ids = ['subject_' + str(subject_id) for subject_id in
range(subject_id, subject_id+len(subjects))]
data_types = [data_type for i in range(len(subjects))]
subject_info = {'subjects': subjects_name,
'subjects_id': subj_ids,
'data_type': data_types}
subject_info = pd.DataFrame.from_dict(subject_info)
subject_info = subject_info.set_index('subjects')
if os.path.exists(csv_file):
with open(csv_file, 'a') as f:
subject_info.to_csv(f, index='subjects', header=False)
else:
subject_info.to_csv(csv_file, index='subjects')
def save_the_data(data_dir_all, subdir, data_dirs, subj_id_init,
remove_previous=False):
# data_dir_all : where to save the data
# data_dirs : paths of the saved data
# remove_previous : clean up the data_dir_all directory before saving new
# files
sbj_id = 0
# initialize the files in the data_dir_all
data_dir_all_subdir = os.path.join(data_dir_all, subdir)
all_X_file = os.path.join(data_dir_all_subdir, 'X.csv.gz')
# remove previous save directory and create a clean one
if os.path.isdir(data_dir_all) and remove_previous:
print('removing contents from ' + data_dir_all)
shutil.rmtree(data_dir_all)
if not os.path.exists(data_dir_all):
os.makedirs(data_dir_all)
if not os.path.exists(data_dir_all_subdir):
os.makedirs(data_dir_all_subdir)
for idx, subject_path in enumerate(data_dirs):
# check if all the necessary files are present
subject_info = subject_path.split('_')
subject_name = subject_info[-3]
subject_name_id = 'subject_' + str(subj_id_init + idx)
# make sure that all the necessary files are in the directory
labels_exist = os.path.exists(os.path.join(subject_path,
subject_name + '_labels.npz'))
target_exists = os.path.exists(os.path.join(subject_path,
'target.npz'))
X_exists = os.path.exists(os.path.join(subject_path, 'X.csv'))
lf_exists = os.path.exists(os.path.join(subject_path,
'lead_field.npz'))
if not (labels_exist and target_exists and X_exists and lf_exists):
print(f'skipping {subject_path}.'
'not all the necessary files are present')
continue
print(f'adding subject {subject_name} as {subject_name_id}')
subject_data = pd.read_csv(os.path.join(subject_path, 'X.csv'))
subject_data['subject'] = subject_name_id
target_subject = sparse.load_npz(os.path.join(subject_path,
'target.npz'))
if sbj_id == 0:
# create new .csv file
subject_data.to_csv(all_X_file, header=True, index=False,
compression='gzip')
target_all = target_subject
else:
# append the data
subject_data.to_csv(all_X_file, mode='a', header=False,
index=False, compression='gzip')
target_all = sparse.vstack((target_all, target_subject))
# shutil.copyfile(os.path.join(subject_path, 'labels.pickle'),
# os.path.join(data_dir_all,
# subject_name + '_labels.pickle')
# )
shutil.copyfile(os.path.join(subject_path, 'lead_field.npz'),
os.path.join(data_dir_all,
subject_name_id + '_L.npz')
)
# uncomment if you want to also save labels
# shutil.copyfile(os.path.join(subject_path,
# subject_name + '_labels.npz'),
# os.path.join(data_dir_all,
# subject_name_id + '_labels.npz')
# )
sbj_id += 1
if len(data_dirs):
# save the target
save_npz(os.path.join(data_dir_all_subdir, 'target.npz'), target_all)
print(f'{target_all.shape[0]} samples from {sbj_id} subjects'
f' were saved in the {data_dir_all} + {subdir}')
else:
print(f'Did not find any files in {data_dir}')
if len(data_dirs) % divide_to != 0:
print('here we are using equal number of data files in each directory.'
f'{len(data_dirs) % divide_to} of the files will not be used')
data_dirs = data_dirs[:-len(data_dirs) % divide_to]
if divide_to == 4:
# divide all the data to public/train, public/test, private/test and
# private/train equally
n_files = int(len(data_dirs) / 4)
# public/train
public_train_dir = os.path.join(data_dir_save, 'public')
save_subject_coords(subject_id_start, data_dirs[n_files:], 'public-train')
save_the_data(public_train_dir, subdir='train',
data_dirs=data_dirs[:n_files],
subj_id_init=subject_id_start)
# public/test
public_test_dir = os.path.join(data_dir_save, 'public')
save_subject_coords(subject_id_start, data_dirs[n_files:], 'public-test')
save_the_data(public_test_dir, subdir='test',
data_dirs=data_dirs[n_files:n_files*2],
subj_id_init=subject_id_start+n_files)
# private/train
private_train_dir = os.path.join(data_dir_save, 'private')
save_subject_coords(subject_id_start, data_dirs[n_files:], 'private-train')
save_the_data(private_train_dir, subdir='train',
data_dirs=data_dirs[n_files*2:n_files*3],
subj_id_init=subject_id_start+n_files*2)
# private/test
private_test_dir = os.path.join(data_dir_save, 'private')
save_subject_coords(subject_id_start, data_dirs[n_files:], 'private-test')
save_the_data(private_test_dir, subdir='test',
data_dirs=data_dirs[n_files*3:],
subj_id_init=subject_id_start+n_files*3)
elif divide_to == 2:
# divide all the data to public and private equally
n_files = int(len(data_dirs) / 2)
# public/train
public_train_dir = os.path.join(data_dir_save, 'public')
save_subject_coords(subject_id_start, data_dirs[:n_files], 'public')
save_the_data(public_train_dir, subdir,
data_dirs=data_dirs[:n_files],
subj_id_init=subject_id_start)
# public/test
public_test_dir = os.path.join(data_dir_save, 'private')
save_subject_coords(subject_id_start, data_dirs[n_files:], 'private')
save_the_data(public_test_dir, subdir,
data_dirs=data_dirs[n_files:],
subj_id_init=subject_id_start)
elif divide_to == 1:
save_the_data(data_dir_save, data_dirs, subj_id_init=subject_id_start)