def process_data_links(line, r, s):
# 输入链路序列line, 每个网络包含节点数量r,跳步数量s
# 输出切分好的链路序列和节点数量
arr = []
for data in line:
data1 = data.strip('\n')
# a, b, _ = data1.split('，') # 这里不同的数据即要做不同的处理
a, b, c, d = data1.split(' ')
arr.append([a, b])
# list->array
arr_2 = np.array(arr, int)
node_num = np.max(arr_2)
# print("共变化了多少时间点：", len(arr_2))
data = [] # 最后的数据存放的矩阵
i, j = 0, r # i表示窗口头部坐标，j表示窗口尾部坐标
while j < len(arr_2):
links = arr_2[i:j, :]
data.append(links) # 加入最后存放数据的矩阵
i, j = i + s, j + s # 窗口滑动步长s
# 处理最后不足窗口大小的时间点数量
links = arr_2[len(arr_2) - r:len(arr_2), :]
data.append(links)
data = np.array(data)
# print(data.shape)
return data, node_num
def __init__(self, data_root, train=True, transform=None):
"""
Args:
1. image folder
2. data name, label list
3. if train, loading data from train folder, or test folder
4.
"""
self.root = data_root
self.transform = transform
self.train = train
if train:
data_list = os.path.join(self.root, 'train.csv')
else:
data_list = os.path.join(self.root, 'test.csv')
with open(data_list) as fin:
data_list = fin.readlines()
self.img_paths = []
self.img_labels = []
self.n_data = 0
for data in data_list[1:]:
data = data.strip('\n').split(',')
self.img_paths.append(data[0])
self.img_labels.append(data[1])
self.n_data += 1
def process_data_matrices(line, r, s):
arr = []
for data in line:
data1 = data.strip('\n')
# a, b, _ = data1.split(',') # 这里不同的数据即要做不同的处理
a, b, c, d = data1.split(' ')
arr.append([a, b])
# list->array
arr_2 = np.array(arr, int)
node_num = np.max(arr_2)
# print("共变化了多少时间点：", len(arr_2))
data = [] # 最后的数据存放的矩阵
i, j = 0, r # i表示窗口头部坐标，j表示窗口尾部坐标
while j < len(arr_2):
matric = np.eye(node_num).astype('int32') # 这是每一份数据的矩阵，np.eye()表示增加了自环
for k in range(i, j):
# 每一个滑动窗口内连接的边都将矩阵相应位置变1
matric[arr_2[k][0] - 1, arr_2[k][1] - 1] += 1
matric[arr_2[k][1] - 1, arr_2[k][0] - 1] += 1
data.append(matric) # 加入最后存放数据的矩阵
i, j = i + s, j + s # 窗口滑动步长s
# 处理最后不足窗口大小的时间点数量
matric = np.eye(node_num).astype('int32')
for k in range(i, len(arr_2)):
matric[arr_2[k][0] - 1, arr_2[k][1] - 1] += 1
matric[arr_2[k][1] - 1, arr_2[k][0] - 1] += 1
data.append(matric)
data = np.array(data)
return data, node_num
def __init__(self, dataset_name, train=True, transform=None, class_num=0):
"""
Args:
1. image folder
2. data name, label list
3. if train, loading data from train folder, or test folder
4.
"""
self.dataset_name = dataset_name
self.transform = transform
self.train = train
if train:
data_list = os.path.join(constant.data_root, dataset_name,
'{}_train.csv'.format(dataset_name))
else:
data_list = os.path.join(constant.data_root, dataset_name,
'{}_test.csv'.format(dataset_name))
with open(data_list) as fin:
data_list = fin.readlines()
self.img_paths = []
self.img_labels = []
self.n_data = 0
self.class_cnt = dict()
for data in data_list[1:]:
data = data.strip('\n').split(',')
img_path, label = data[0], data[1]
self.img_paths.append(img_path)
self.img_labels.append(label)
self.n_data += 1
if label in self.class_cnt.keys():
self.class_cnt[label] += 1
else:
self.class_cnt[label] = 1
self.classes = [
name for name in os.listdir(
os.path.join(constant.data_root, dataset_name))
if 'csv' not in name
]
self.classes = sorted(self.classes, key=lambda s: s.lower())
if class_num > 0:
pick_class_code = list(range(class_num))
temp_img_path = []
temp_img_label = []
for (img_path, label) in zip(self.img_paths, self.img_labels):
if int(label) in pick_class_code:
temp_img_path.append(img_path)
temp_img_label.append(label)
self.img_paths = temp_img_path
self.img_labels = temp_img_label
self.classes = self.classes[:class_num]
def __init__(self, image_root, landmark_file, transform):
super(ImageFeatureFolder, self).__init__(root=image_root,
transform=transform)
with open(landmark_file, 'r') as f:
data = f.read()
data = data.strip().split('\n')
self.attrs = torch.FloatTensor(
[list(map(float,
line.split()[1:])) for line in data[2:]])
def preprocess(self, data):
# pdb.set_trace()
values = data.strip().split()
labels = float(values[0])
indexs = [
int(v.split(':')[0]) - sum(self.feature_sizes[:i])
for i, v in enumerate(values[1:])
]
socres = [int(v.split(':')[1]) for v in values[1:]]
# pdb.set_trace()
return labels, torch.LongTensor(indexs), torch.FloatTensor(socres)
def handle_data(self, data):
if self.in_table:
if data == 'Image Not Found':
self.current_img = None
elif self.current_tag == 'a':
img_id = data.split('/')[-2]
img_id = os.path.join(self.root, img_id + '_*.jpg')
img_id = glob.glob(img_id)[0]
self.current_img = img_id
self.annotations[img_id] = []
elif self.current_tag == 'li' and self.current_img:
img_id = self.current_img
self.annotations[img_id].append(data.strip())