def __init__(self, root, list_file, train, transform, transform_att,
input_size):
'''
Args:
root: (str) ditectory to images.
list_file: (str) path to index file.
train: (boolean) train or test.
transform: ([transforms]) image transforms.
input_size: (int) model input size.
'''
self.root = root
self.train = train
self.transform = transform
self.transform_att = transform_att
self.input_size = input_size
self.fnames = []
self.boxes = []
self.ids = []
self.get_att = seg_attention(input_size)
self.encoder = DataEncoder()
file_list = csv.reader(open(list_file, 'r'))
file_list = list(file_list)
for csv_content in file_list:
self.fnames.append(csv_content[0])
self.ids.append(int(csv_content[1]))
self.thresh = torch.nn.Hardtanh(min_val=0, max_val=1)
def __init__(self, root, train, transform, transform_att, input_size):
'''
Args:
root: (str) ditectory to images.
list_file: (str) path to index file.
train: (boolean) train or test.
transform: ([transforms]) image transforms.
input_size: (int) model input size.
'''
self.root = root
self.train = train
self.transform = transform
self.transform_att = transform_att
self.input_size = input_size
self.get_att = seg_attention(input_size)
self.thresh = torch.nn.Hardtanh(min_val=0, max_val=1)
fnames = []
self.boxes = []
ids = []
self.ids_list = list(range(2874))
self.im_name_list = []
self.encoder = DataEncoder()
file_list = os.listdir(root)
# 2874
self.im_name_train = file_list
self.ids_train = list(range(len(file_list)))
def __init__(self, root, list_file, train, transform, transform_att, input_size):
'''
Args:
root: (str) ditectory to images.
list_file: (str) path to index file.
train: (boolean) train or test.
transform: ([transforms]) image transforms.
input_size: (int) model input size.
'''
self.root = root
self.train = train
self.transform = transform
self.transform_att = transform_att
self.input_size = input_size
self.get_att = seg_attention(input_size)
self.thresh = torch.nn.Hardtanh(min_val=0, max_val=1)
fnames = []
self.boxes = []
ids = []
self.ids_list = list(range(2874))
self.im_name_list = []
self.encoder = DataEncoder()
file_list = csv.reader(open(list_file,'r'))
file_list = list(file_list)
# 2874
for content_counter in range(len(file_list)):
fnames.append(file_list[content_counter][0])
ids.append(int(file_list[content_counter][1]))
'''
for id_counter in range(2874):
seq_num = ids.index(id_counter)
self.im_name_list.append(fnames[seq_num])
del(ids[seq_num])
del(fnames[seq_num])
self.im_name_valid = fnames[:400]
self.im_name_train = fnames[400:]+self.im_name_list
self.ids_valid = ids[:400]
self.ids_train = ids[400:]+self.ids_list
'''
self.im_name_train = fnames
self.ids_train = ids
def __init__(self,
root,
list_file,
input_size,
eval_num,
addition_root=None,
addition_list=None,
transform=transform,
transform_att=transform_pure,
att_flag=False,
flip_flag=True,
random_crop_flag=False,
center_crop_flag=True,
align_flag=True,
state = 'train'):
self.root = root
self.list_file = list_file
if addition_root is not None:
self.addition_root = addition_root
self.addition_list = addition_list
self.input_size = input_size
self.eval_num = eval_num
self.att_flag = att_flag
self.align_flag = align_flag
self.transform = transform
self.transform_att = transform_att
self.random_crop_flag = random_crop_flag
self.center_crop_flag = center_crop_flag
self.flip_flag = flip_flag
fnames = []
ids = []
self.encoder = DataEncoder()
file_list = list(csv.reader(open(list_file, 'r')))
for content_counter in file_list:
fnames.append(os.path.join(self.root, content_counter[0]))
ids.append(int(content_counter[1]))
im_name_list = []
for id_counter in range(2874):
seq_num = ids.index(id_counter)
im_name_list.append(fnames[seq_num])
del(ids[seq_num])
del(fnames[seq_num])
ids_list = list(range(2874))
self.im_name_valid = fnames[:self.eval_num]
self.im_name_train = fnames[self.eval_num:]+im_name_list
self.ids_valid = ids[:self.eval_num]
self.ids_train = ids[self.eval_num:]+ids_list
if self.align_flag:
self.align_detect = dlib.get_frontal_face_detector()
predicter_path = "./model/shape_predictor_5_face_landmarks.dat"
self.sp = dlib.shape_predictor(predicter_path)
if addition_root is not None:
self.addition_root = addition_root
self.addition_list = addition_list
add_csv = list(csv.reader(open(self.addition_list, 'r')))
self.add_id = []
self.add_img = []
for img_list in add_csv:
self.add_id.append(int(img_list[1]))
folder_dir = re.sub(r'_\d\d\d\d.jpg', '', img_list[0])
img_dir = os.path.join(os.path.join(self.addition_root, folder_dir), img_list[0])
self.add_img.append(img_dir)
if state == 'train':
self.excuse_list = self.im_name_train
self.excuse_ids = self.ids_train
else:
self.excuse_list = self.im_name_valid
self.excuse_ids = self.ids_valid
if addition_root is not None:
self.excuse_list += self.add_img
self.excuse_ids += self.add_id
self.thresh = torch.nn.Hardtanh(min_val=0, max_val=1)
self.get_att = seg_attention(input_size)