def get_raw_one_slice(subject_num, slice_num, folder_name='./train_data/', label=False):
from src.util import read_data
ret = np.empty(( _width, _height), dtype =np.int16)
if subject_num < 10:
subject_name = 'Subject0' + str(subject_num)
else:
subject_name = 'Subject' + str(subject_num)
if label is True:
fn = folder_name + subject_name + '/' + subject_name + '_Label.nii'
else:
fn = folder_name + subject_name + '/' + subject_name + '.nii'
arr = read_data(fn).get_data()
ret = arr[slice_num]
return ret
def get_raw_slices(total_slice_num = _slice_num, slice_nums=[], subject_nums=[], folder_name='./train_data/', label=False):
from src.util import read_data
if subject_nums == []:
subject_nums = range(1, _subject_number + 1)
ret = np.empty((total_slice_num, _width, _height), dtype =np.int16)
idx = 0
for i in subject_nums:
if i < 10:
subject_name = 'Subject0' + str(i)
else:
subject_name = 'Subject' + str(i)
if label is True:
fn = folder_name + subject_name + '/' + subject_name + '_Label.nii'
else:
fn = folder_name + subject_name + '/' + subject_name + '.nii'
arr = read_data(fn).get_data()
if slice_nums == []:
begin = arr.shape[0]/2 - 2
end = arr.shape[0]/2 + 2
slice_nums = np.arange(begin, end + 1)
# this this for some slices with different width and height
offset = slice_nums.shape[0]
if arr.shape[-1] != _height:
diff_y = _height - arr.shape[1]
diff_x = _width - arr.shape[2]
ret[idx: idx + offset, :, :] =np.lib.pad( arr[slice_nums], ((0, 0), (diff_y, 0), (diff_x,0)), 'minimum')
else:
ret[idx:idx+offset, :, :] = arr[slice_nums]
idx = idx + offset
return ret
print(logs)
self.losses.append(logs.get('loss'))
if __name__ == '__main__':
train_file_filter = "train_*.txt"
test_file_filter = "test_*.txt"
seq_length = 150
embedding_dim = 50 # 単語ベクトルの次元数
lstm_units = 50 # LSTMの隠れ状態ベクトルの次元数
epoch_size = 5
batch_size = 20
# read source data
label_dict = util.create_label('./train_master.tsv')
sentences_train, labels, train_files = util.read_data(
"./train/", train_file_filter, label_dict)
sentences_test, labels_test, test_files = util.read_data(
"./test/", test_file_filter, None)
train, test, seq_length, vocab_size = util.create_train_data(
sentences_train, sentences_test, seq_length)
X_train, X_test, y_train, y_test = train_test_split(
train, labels, test_size=0.15, random_state=random.randint(0, 100))
# create model
model = model.create_bidirectional(vocab_size=vocab_size,
embedding_dim=embedding_dim,
seq_length=seq_length,
lstm_units=lstm_units)
#model = model.create_3layer(seq_length=seq_length)
import sys dir = os.getcwd() sys.path.append(dir) result_dir = dir + '/result/test_1/' test_data_dir = dir + '/train_data/' import numpy as np from src.asm_util import get_bulk_points slice_num = 5 v = np.empty((7, slice_num, 16), dtype=np.int) v[0], v[1], v[2], v[3], v[4], v[5], v[6] = get_bulk_points() from src.util import read_data subject = 'Subject01' label_fn = test_data_dir + subject + '/' + subject + '_Label.nii' label_data = read_data(label_fn).get_data() total_slice_num = label_data.shape[0] from src.asm_util import get_profile_feature_one_vertebra from src.asm_util import get_raw_slices vertebra_number = 6 I = get_raw_slices() #L = get_raw_slices(vertebra_number, label=True) #from src.util import show_volume #show_volume(I,75) _K = 1 _sigmas = np.array([2, 3]) mid = total_slice_num/2 label = label_data[mid-2 : mid+3] samplepoints, intensities, labels, features= get_profile_feature_one_vertebra(v[vertebra_number], I, _k, label, sigmas = _sigmas)
range_start = 15
range_end = range_start + 1
fnum = range(range_start, range_end)
from src.util import read_data
label_data={}
data = {}
subject = ''
for i in fnum:
if i < 10:
subject = 'Subject0' + str(i)
else:
subject = 'Subject' + str(i)
label_fn = test_data_dir + subject + '/' + subject + '_Label.nii'
label_data[i] = read_data(label_fn)
fn = test_data_dir + subject + '/' + subject + '.nii'
data[i]=read_data(fn)
#sumsum = 0
#num = 0
#for i in fnum:
# idx = np.where(label_data[i].get_data() == 1)
# num += np.array(idx[0]).size
# sumsum += np.sum(data[i].get_data()[idx])
#
#average = sumsum/num
#================= get the location of the center ==================
# goto slice 20
#slice_num = 20
