def image_styletransfer():
if request.method == 'POST':
if request.form.get('filename') is None:
# 接收图片并保存
im = request.files['image']
filename = save_file(im)
im2 = request.files['style_image']
filename2 = save_file(im2)
style_name = im2.filename.split('.')[0]
print(style_name)
if filename is None:
return render_template('error.html', msg='不支持的文件格式!')
else:
filename = request.form.get('filename')
# 滤波
res_name = img_style_transfer(filename, filename2)
# 返回结果
return redirect(
url_for('filter_bp.filter_result',
mode=style_name,
back='style_transfer',
filename=filename,
res_name=res_name,
flag='style_transfer'))
else:
return render_template('filter/style_transfer.html')
def create_triples(relations, house_objects, visualgenome_path):
''' Create RDF triples of the relations in two formats: labels and URIs '''
house_object_uris = set(house_objects.values())
house_object_names = set(house_objects.keys())
lemmadb = map_lemmadb(lowercase=True)
relations_with_uris = []
triple_labels = []
triple_uris = []
for object_name, category, attribute in relations:
if object_name in lemmadb and attribute in lemmadb: # objects and attributes must to have URIs
object_uri = create_uri(lemmadb[object_name])
object_name = object_name.split('.')[0]
if object_uri in house_object_uris or object_name in house_object_names:
relations_with_uris.append((object_name, category, attribute))
triple_labels.append(
'<%s> <%s> <%s>' %
(object_name, category, attribute.split('.')[0]))
category_uri = 'http://ns.inria.fr/deko/ontology/deko.owl#' + category
triple_uris.append(
'<%s> <%s> <%s>' %
(object_uri, category_uri, create_uri(lemmadb[attribute])))
calculate_statistics(relations_with_uris, visualgenome_path)
save_file(join(visualgenome_path, 'selected_triples.nt'), triple_uris)
save_file(join(visualgenome_path, 'selected_triples_label.nt'),
triple_labels)
logging.info('Total valid relations with URIs: %s' %
len(relations_with_uris))
def calculate_statistics(object_relations, concepnet_parsed_path):
''' Calculate statistics of the dataset '''
relation_counter = {}
object_counter = set()
for object1, relation, object2 in object_relations:
if relation not in relation_counter:
relation_counter[relation] = 0
relation_counter[relation] += 1
object_counter.add(object1)
object_counter.add(object2)
table = PrettyTable()
table.field_names = ['Item', 'Amount']
table.align['Item'] = 'l'
table.align['Amount'] = 'r'
table.add_row(['Objects', len(object_counter)])
table.add_row(
['Relations (object-object)',
sum(relation_counter.values())])
for relation in relation_counter.keys():
table.add_row(['Relation "%s"' % relation, relation_counter[relation]])
print(table)
save_file(join(concepnet_parsed_path, 'statistics_dataset.txt'),
[table.get_string()])
def calculate_statistics(object_relations, visualgenome_path):
''' Calculate statistics of the dataset '''
category_counter = {}
object_counter = set()
attribute_counter = []
for object, category, attributte in object_relations:
if category not in category_counter:
category_counter[category] = 0
category_counter[category] += 1
object_counter.add(object)
attribute_counter.append(attributte)
table = PrettyTable()
table.field_names = ['Item', 'Amount']
table.align['Item'] = 'l'
table.align['Amount'] = 'r'
total_objects = len(object_counter)
table.add_row(['Objects', total_objects])
table.add_row(['Attributes', len(set(attribute_counter))])
table.add_row(['Relations (object-attribute)', len(object_relations)])
total_objects = float(total_objects)
for category in sorted(category_counter.keys()):
table.add_row(
['Relations "%s"' % category, category_counter[category]])
table.add_row([
'Avg. attributes by object',
round(len(attribute_counter) / total_objects, 2)
])
for category in sorted(category_counter.keys()):
table.add_row([
'Avg. relations "%s" by object' % category,
round(category_counter[category] / total_objects, 2)
])
print(table)
save_file(join(visualgenome_path, 'statistics_dataset.txt'),
[table.get_string()])
table.clear_rows()
table.field_names = ['Attribute', 'Frequency']
for attribute, frequency in sorted(Counter(attribute_counter).items(),
key=lambda x: x[1],
reverse=True):
table.add_row([attribute, frequency])
save_file(join(visualgenome_path, 'statistics_attributes.txt'),
[table.get_string()])
def init_train(self, train_path, data_path="store/train_data.pkl"):
"""初始化数据"""
try:
print("开始载入初始化数据....", file=sys.stderr)
train_data = load_file(data_path)
except BaseException:
train_data = dict()
for line in process_data.read_file(filename=train_path):
arr = line.split('\t')
train_data.setdefault(int(arr[0]), set())
train_data[int(arr[0])].add(int(arr[1]))
save_file(data_path, train_data)
return train_data
def train(self, sim_matrix_path="store/buy_user_sim.pkl"):
print("开始训练模型", file=sys.stderr)
try:
print("开始载入用户协同矩阵....", file=sys.stderr)
self.user_sim_matrix = load_file(sim_matrix_path)
print("载入协同过滤矩阵完成", file=sys.stderr)
except BaseException:
print("载入用户协同过滤矩阵失败，重新计算协同过滤矩阵", file=sys.stderr)
# 计算用户协同矩阵
self.user_sim_matrix = self.user_similarity()
print("开始保存协同过滤矩阵", file=sys.stderr)
save_file(sim_matrix_path, self.user_sim_matrix)
print("保存协同过滤矩阵完成", file=sys.stderr)
def image_filter():
if request.method == 'POST':
if request.form.get('filename') is None:
# 接收图片并保存
im = request.files['image']
filename = save_file(im)
if filename is None:
return render_template('error.html', msg='不支持的文件格式!')
else:
filename = request.form.get('filename')
# 获取滤波方式
mode = request.form.get('mode', None)
# 滤波
res_name, olddata = img_filter(request.form, filename, mode=mode)
# 返回结果
print(filename)
return redirect(
url_for('filter_bp.filter_result',
mode=mode,
back='filter',
filename=filename,
res_name=res_name,
flag='filter',
olddata=olddata))
else:
return render_template('filter/filter.html')
def image_enhance():
if request.method == 'POST':
if request.form.get('filename') is None:
im = request.files['image']
# 保存图像
filename = save_file(im)
if filename is None:
return render_template('error.html', msg='不支持的文件格式!')
else:
filename = request.form.get('filename')
enhance_type = request.form.get('enhance', None)
factor_select = request.form.get('factor-select', None)
factor_custom = request.form.get('factor-custom', None)
# 设置增强因子
if factor_custom != '':
factor = float(factor_custom)
else:
factor = float(factor_select)
# 图像增强
res_name = img_enhance(filename, enhance_type, factor)
return redirect(url_for('filter_bp.filter_result', mode=enhance_type, back='enhance', filename=filename, res_name=res_name,
flag='enhance', olddata=factor))
else:
return render_template('filter/enhance.html')
def image_kernel():
if request.method == 'POST':
if request.form.get('filename') is None:
# 保存图像
im = request.files.get('image', None)
# 保存图像
filename = save_file(im)
if filename is None:
return render_template('error.html', msg='不支持的文件格式!')
else:
filename = request.form.get('filename')
# 获取卷积核信息
size = request.form.get('size', None)
kernel_size, kernel = get_kernel(size)
scale = request.form.get('scale', None)
offset = request.form.get('offset', None)
if kernel is None:
return redirect(url_for('filter_bp.image_kernel'))
# 处理图像
res_name = img_kernel(filename, kernel_size, kernel, scale, offset)
olddata = '#'.join(str(e) for e in kernel) + '#' + scale + '#' + offset
return redirect(url_for('filter_bp.filter_result', mode='Custom Kernel', back='kernel', filename=filename,
res_name=res_name, olddata=olddata, flag='kernel-' + size))
else:
return render_template('filter/kernel.html')
def train(self, origin_data, sim_matrix_path="store/user_sim.pkl"):
"""训练模型
@param origin_data: 原始数据
@param sim_matrix_path: 协同矩阵保存的路径
"""
self.origin_data = origin_data
# 初始化训练集
self._init_train(origin_data)
print("开始训练模型", file=sys.stderr)
try:
print("开始载入用户协同矩阵....", file=sys.stderr)
self.user_sim_matrix = load_file(sim_matrix_path)
print("载入协同过滤矩阵完成", file=sys.stderr)
except BaseException:
print("载入用户协同过滤矩阵失败，重新计算协同过滤矩阵", file=sys.stderr)
# 计算用户协同矩阵
self.user_sim_matrix = self.user_similarity()
print("开始保存协同过滤矩阵", file=sys.stderr)
save_file(sim_matrix_path, self.user_sim_matrix)
print("保存协同过滤矩阵完成", file=sys.stderr)
def water_mask_img():
if request.method == 'POST':
mask_type = request.form.get('type')
file1 = request.form.get('file1')
file2 = request.form.get("file2")
if file1 is None:
# 获取原始图像数据
im1 = request.files['image1']
file1 = save_file(im1)
olddata = ''
if mask_type == 'logo':
if file2 is None:
# Logo水印
logo = request.files['image2']
file2 = save_file(logo)
res_filename = water_mask(file1, file2)
else:
# 获取文字水印内容
text = request.form.get('text')
color = request.form.get('color')
font_size = request.form.get('font_size')
x = request.form.get('x')
y = request.form.get('y')
res_filename, olddata = text_mask(file1, text, color, font_size,
(x, y))
args = {
'back': 'tools_bp.water_mask_img',
'res_filename': res_filename,
'file1': file1,
'file2': file2,
'time': time(),
'olddata': olddata
}
return redirect(url_for('tools_bp.water_mask_result', **args))
else:
return render_template('tools/water_mask.html')
def calculate_statistics(object_relations, output_path):
''' Calculate statistics of the dataset '''
ftype_counter = set()
felement_counter = set()
ffiller_counter = set()
for triple in object_relations:
ffiller, felement, ftype = re.match('<(.+)> <(.+)> <(.+)>', triple).groups()
ftype_counter.add(ftype)
felement_counter.add(felement)
ffiller_counter.add(ffiller)
table = PrettyTable()
table.field_names = ['Item', 'Amount']
table.align['Item'] = 'l'
table.align['Amount'] = 'r'
table.add_row(['Frame types', len(ftype_counter)])
table.add_row(['Frame elements', len(felement_counter)])
table.add_row(['Frame elements fillers', len(ffiller_counter)])
table.add_row(['Relations', len(object_relations)])
print(table)
save_file(join(output_path,'statistics_dataset.txt'), [table.get_string()])
def blend_img():
if request.method == "POST":
file1 = request.form.get('file1', None)
file2 = request.form.get('file2', None)
if file1 is None or file2 is None:
# 获取图像并保存
im1 = request.files['image1']
im2 = request.files['image2']
file1 = save_file(im1)
file2 = save_file(im2)
# 获取alpha
alpha = request.form['alpha']
res_filename = overlay(file1, file2, float(alpha))
args = {
'back': 'tools_bp.blend_img',
'file1': file1,
'file2': file2,
'res_filename': res_filename,
'time': time(),
'olddata': alpha
}
return redirect(url_for('tools_bp.blend_result', **args))
else:
return render_template('tools/blend.html')
def select_relations(conceptnet_raw_path, concepnet_parsed_path,
house_objects_path): #
''' Select some relations from Conceptnet JSON files '''
validated_relations = validate_relations(conceptnet_raw_path)
wn31db = map_wn31db()
relations_with_uris = []
objects_with_uris = {
v.replace(' ', '_'): k['dbpedia_uri']
for v, k in load_json(house_objects_path).items()
}
triple_labels = []
triple_uris = []
for object1, relation, object2 in validated_relations:
if object2 not in objects_with_uris:
objects_with_uris[object2] = to_dbpedia(
get_uri('/c/en/' + object2, 10), wn31db)
for object1, relation, object2 in validated_relations:
if objects_with_uris[
object2]: # object2 must to have an URI (by default object1 has an URI)
relation_uri = 'http://ns.inria.fr/deko/ontology/deko.owl#' + relation
relations_with_uris.append((objects_with_uris[object1], relation,
objects_with_uris[object2]))
triple_uris.append('<%s> <%s> <%s>' %
(objects_with_uris[object1], relation_uri,
objects_with_uris[object2]))
triple_labels.append('<%s> <%s> <%s>' %
(object1, relation, object2))
calculate_statistics(relations_with_uris, concepnet_parsed_path)
save_file(join(concepnet_parsed_path, 'selected_triples.nt'), triple_uris)
save_file(join(concepnet_parsed_path, 'selected_triples_label.nt'),
triple_labels)
logging.info('Total valid relations with URIs: %s' %
len(relations_with_uris))
def resize_img():
if request.method == 'POST':
filename = request.form.get('filename')
if filename is None:
img = request.files['image']
filename = save_file(img)
prop = float(request.form.get('proportion'))
if prop < 0.1:
raise Exception('比例系数不合法!')
res_filename = resize(filename, proportion=prop)
args = {
'back': 'tools_bp.resize_img',
'filename': filename,
'res_filename': res_filename,
'olddata': prop
}
return redirect(url_for('tools_bp.tools_result', **args))
else:
return render_template('tools/resize.html')
def mosaic_img():
if request.method == 'POST':
# 获取文件名
filename = request.form.get('filename')
if filename is None:
im = request.files['image']
filename = save_file(im)
# 获取参数
startx = request.form.get('startx')
starty = request.form.get('starty')
endx = request.form.get('endx')
endy = request.form.get('endy')
granularity = int(request.form.get('g'))
res_filename, olddata = mosaic(filename, [startx, starty],
[endx, endy], granularity)
args = {
'back': 'tools_bp.mosaic_img',
'filename': filename,
'res_filename': res_filename,
'olddata': olddata
}
return redirect(url_for('tools_bp.tools_result', **args))
else:
return render_template('tools/mosaic.html')
def select_relations(frame_parsed_path, house_objects_path):
''' Select unique relations of frames about house's object '''
house_objects = {v.replace(' ', '_'):k['dbpedia_uri'] for v,k in load_json(house_objects_path).items()}
house_object_uris = set(house_objects.values())
house_object_names = set(house_objects.keys())
frame_instances = load_json(join(frame_parsed_path, 'frame_instances.json'))
netlemma = map_netlemma()
wn31db = map_wn31db()
triple_uris = []
triple_labels = []
for frame_id in frame_instances.keys():
valid_frame = False
frame = frame_instances[frame_id]
frame_uris, frame_labels = create_triples(frame['type'], frame['elements'], wn31db, netlemma)
for i in range(len(frame_uris)):
object_uri = re.match('<(.+)> <(.+)> <(.+)>', frame_uris[i]).group(1)
if object_uri in house_object_uris and frame_uris[i] not in triple_uris:
triple_uris.append(frame_uris[i])
triple_labels.append(frame_labels[i])
valid_frame = True
else:
object_name = re.match('<(.+)> <(.+)> <(.+)>', frame_labels[i]).group(1)
if object_name in house_object_names and frame_uris[i] not in triple_uris:
triple_uris.append(frame_uris[i])
triple_labels.append(frame_labels[i])
valid_frame = True
if not valid_frame:
del frame_instances[frame_id]
calculate_statistics(triple_uris, frame_parsed_path)
save_file(join(frame_parsed_path, 'selected_triples.nt'), triple_uris)
save_file(join(frame_parsed_path, 'selected_triples_label.nt'), triple_labels)
save_file(join(frame_parsed_path, 'selected_verbalized.txt'), [verbalize_frame(f['type'], f['elements'].items(), netlemma) for f in frame_instances.values()])
logging.info('Total valid relations with URIs: %s' % len(triple_uris))
def user_similarity(self):
"""建立用户的协同过滤矩阵"""
user_list = self.train_data.keys()
for user in user_list:
items = self.train_data.get(user)
return
# 建立用户倒排表
item_user = dict()
item_user_path = "store/item_user.pkl"
try:
print("开始载入用户倒排表....", file=sys.stderr)
item_user = load_file(item_user_path)
except BaseException:
for user, items in self.train_data.items():
print('user:'******'item:' + str(item))
for u in users:
N[u] += 1
save_file(N_path, N)
try:
print("开始载入用户协同过滤矩阵....", file=sys.stderr)
user_sim_matrix = load_file(user_sim_matrix_path)
except BaseException:
i = 1
print(len(item_user))
for item, users in item_user.items():
print(i)
i += 1
# print('item:' + str(item))
for u in users:
for v in users:
if u == v:
continue
# user_sim_matrix.setdefault(u, defaultdict(int))
# user_sim_matrix[u][v] += 1
save_file(user_sim_matrix_path, user_sim_matrix)
# print(N)
# # 计算相关度
# for u, related_users in user_sim_matrix.items():
# for v, con_items_count in related_users.items():
# user_sim_matrix[u][v] = con_items_count / math.sqrt(N[u] * N[v])
return user_sim_matrix
