def _prepare_data(self):
all_im_list = dfutil.get_file_list(self._data_dir, '.jpg')
class_dict = {}
class_cnt = 0
for im_path in all_im_list:
head, tail = ntpath.split(im_path)
class_label = tail.split('_')[0]
if class_label != '0000' and class_label != '-1':
if class_label in class_dict:
class_id = class_dict[class_label]
self.im_list[class_id].append(im_path)
else:
class_dict[class_label] = class_cnt
try:
self.im_list.append([im_path])
except AttributeError:
self.im_list = [[]]
class_cnt += 1
self._n_class = len(self.im_list)
for idx, class_im_list in enumerate(self.im_list):
self.im_list[idx] = np.array(self.im_list[idx])
self._shuffle_files()
def _load_file_list(self, data_name_list, data_dir_list):
self._file_name_list = [[] for _ in range(len(data_name_list))]
self._file_name_list[0] = get_file_list(data_dir_list[0],
data_name_list[0])
for file_path in self._file_name_list[0]:
drive, path_and_file = os.path.splitdrive(file_path)
path, file = os.path.split(path_and_file)
file_id, file_extension = os.path.splitext(file)
# file_id = re.findall(r'\d+', file)[0]
for idx, (data_name, data_dir) in enumerate(
zip(data_name_list[1:], data_dir_list[1:])):
self._file_name_list[idx + 1].append(
os.path.join(data_dir, file_id + data_name))
for idx, file_list in enumerate(self._file_name_list):
self._file_name_list[idx] = np.array(file_list)
if self._shuffle:
self._suffle_file_list()
n_sample = int(len(self._file_name_list[0]) * self._load_percentage)
for idx, file_list in enumerate(self._file_name_list):
self._file_name_list[idx] = file_list[:n_sample]
print('{} samples in total.'.format(n_sample))
def _load_file_list(self, data_name_list, data_dir_list):
self._file_name_list = []
for data_name, data_dir in zip(data_name_list, data_dir_list):
self._file_name_list.append(get_file_list(data_dir, data_name))
if self._shuffle:
self._suffle_file_list()
def _load_data(self):
im_list = np.array(sorted(dfutil.get_file_list(self._data_dir, 'jpg')))
label_list = [0 for _ in range(len(im_list))]
return im_list, label_list