def detection_image(self,
img,
save_dir,
rotation_angle,
i=None,
file=None,
type='url'):
if type == 'url':
source = i.get('type', '')
img_url = i.get('path', '')
img_time = time.time()
logger.info('[ticket:%s] %s %s detection start.' %
(self.ticket, source, img_url))
detectied_region, img_gray, table_bbox, point_sets = ocr_cv.detect_word(
img,
img_url,
'url',
self.ticket,
companyName=self.companyName,
source=source,
data_dir=save_dir)
logger.info(
'[ticket:%s] %s %s detection end. time cost:%ss' %
(self.ticket, source, img_url, round(time.time() - img_time, 3)))
tmp = {}
tmp['rotation_angle'] = rotation_angle
tmp['detectied_region'] = detectied_region
tmp['img_gray'] = img_gray
tmp['img_name'] = source
tmp['img_url'] = img_url
tmp['id'] = i.get('id', '')
tmp['page'] = i.get('page', '')
tmp['table_bbox'] = table_bbox
tmp['point_sets'] = point_sets
self.detection_list.append(tmp)
else:
img_path = r'%s/%s/%s' % (self.data_dir, file, img)
tmp_img_path = '%s/%s' % (file, img)
detectied_region, img_gray, table_bbox, point_sets = ocr_cv.detect_word(
img, img_path, data_dir=save_dir)
tmp = {}
tmp['rotation_angle'] = rotation_angle
tmp['detectied_region'] = detectied_region
tmp['img_gray'] = img_gray
tmp['img_name'] = file
tmp['img_url'] = ''
tmp['id'] = ''
tmp['page'] = ''
tmp['table_bbox'] = table_bbox
tmp['point_sets'] = point_sets
self.detection_list.append(tmp)
def post_test(post_server=True, server_name=None):
if post_server:
# url='http://apis.cisdi.amiintellect.com/api/cisdi/ml/economic/{}/1234'.format(server_name)
url = 'http://apis.cisdi.amiintellect.com/api/cisdi/ml/economic/{}/1234'.format(
server_name)
else:
url = 'http://8113.204.147.34:8888/api/cisdi/ml/economic/{}/1234'.format(
server_name) # 28095
# data_json=switch_post[server_name]
res = requests.post(
url, # json={"mytext":"from client :lalala"}
json=data_json)
print(res)
if res.ok:
# print('res.json()',res)
logger.info('from server response:{}'.format(
res.json())) #response是post请求的返回值
def image_deal(self):
start_time = time.time()
# 区域检测
detection(self)
logger.info('[ticket:%s] detection_list_len: %d' %
(self.ticket, len(self.detection_list)))
# 区域文字识别
result_e, result_c = recognization(self)
cost_time = time.time() - start_time
# with open('./data/merge_content_test/result_e.txt', 'r', encoding='utf8') as f:
# result_e = eval(f.read())
# with open('./data/merge_content_test/result_c.txt', 'r', encoding='utf8') as f:
# result_c = eval(f.read())
# 将识别的内容按照pdf页码的顺序进行重排,从旋转90和270中的内容中选一个
result_e = result_rearrange(result_e)
result_c = result_rearrange(result_c)
logger.info(
'[ticket:%s] detection and recognization finished. image number:%s, time cost:%ss'
% (self.ticket, self.slice_num, round(cost_time, 3)))
return result_e, result_c
def __init__(self, companyName=None, ticket=None):
start_time = time.time()
self.db = pymysql.connect(host=host_name,
user=user_name,
password=password,
db=database,
charset=charset)
self.cur = self.db.cursor()
# 使用 execute() 方法执行 SQL 查询
self.cur.execute("SELECT VERSION()")
# 使用 fetchone() 方法获取单条数据.
data = self.cur.fetchall()
print("Database version : %s " % data)
self.db.close()
cost_time = time.time() - start_time
logger.info(' time cost:%ss' % (round(cost_time, 3)))
def schedule_func(uuid):
try:
logger.info('开始进入模型，服务端获取数据')
now_time = datetime.now().strftime("%Y-%m-%d-%H-%M-%S")
t1 = time.time()
content = request.get_json(silent=True, force=True)
logger.info('from client type content:{},{}'.format(
type(content), content)) # Do your processing
voyage_id = content["voyage"]
schedule_api(voyage_id)
logger.info('结果已经返回')
# fres={'schedule_position':res}
return jsonify({'code': "0", 'message': "success"})
except Exception as e:
logger.error('出错：{}\n{}'.format(e, traceback.format_exc()))
return jsonify({'data': None, 'code': "1", 'message': "{}".format(e)})
def detection(self):
start_time = time.time()
if self.data_dir:
save_dir = self.data_dir + '/detect_image'
# if os.path.exists(save_dir):
# shutil.rmtree(save_dir)
# if os.path.exists(self.data_dir):
# os.mkdir(save_dir)
if not os.path.exists(save_dir) and os.path.exists(self.data_dir):
os.mkdir(save_dir)
else:
save_dir = ''
# 以url集合的方式处理图片
if self.img_data_dir:
logger.info('[ticket:%s] detection start. by url...' % (self.ticket))
for i in self.img_data_dir:
img_url = i.get('path', '')
suffix = img_url.split('.')[-1]
if suffix not in ['jpg', 'png', 'jpeg']:
continue
# 读取图片
try:
# start_time = time.time()
# logger.info('[ticket:%s] %s read image start.' % (self.ticket, img_url))
image_path = parse.quote(img_url,
encoding='utf8').replace('%3A', ':')
resp = request.urlopen(image_path)
img = np.asarray(bytearray(resp.read()), dtype="uint8")
img = cv2.imdecode(img, cv2.IMREAD_COLOR)
# logger.info('[ticket:%s] %s read image end.time cost:%ss' % (self.ticket, image_path, round(time.time() - start_time, 3)))
except Exception as e:
logger.warning('[ticket:%s] %s error' %
(self.ticket, image_path))
logger.warning(e)
# 判断图片是否需要旋转
if not ocr_rotation.detect_rotation(img, 127):
logger.info('[ticket:%s] %s 旋转图片.' % (self.ticket, img_url))
# 旋转90度，并进行文本检测
img_90 = np.rot90(img)
detection_image(self, img_90, save_dir, i=i, rotation_angle=90)
# 旋转270度，并进行文本检测
img_270 = np.rot90(img, 3)
detection_image(self,
img_270,
save_dir,
i=i,
rotation_angle=270)
detection_image(self, img, save_dir, i=i, rotation_angle=0)
else:
# 图片文本检测和表格检测
detection_image(self, img, save_dir, i=i, rotation_angle=0)
# 处理本地图片
elif not self.ocr_data_files and os.path.exists(r'%s/' % self.data_dir):
# we store unzipped images in company_name's temporary directory
logger.info('[ticket:%s] detection start. by local path. data_dir:%s' %
(self.ticket, self.data_dir))
for file in sorted(os.listdir(r'%s/' % self.data_dir),
key=lambda x: x):
if os.path.isdir((r'%s/' % self.data_dir) + file):
for im in sorted(os.listdir((r'%s/' % self.data_dir) + file),
key=lambda x: x):
suffix = im.split('.')[-1]
if suffix not in ['jpg', 'png', 'jpeg']:
continue
# 读图片
image_path = r'%s/%s/%s' % (self.data_dir, file, im)
img = cv2.imdecode(np.fromfile(image_path, dtype=np.uint8),
-1)
# 判断图片是否需要旋转
if not ocr_rotation.detect_rotation(img, 127):
logger.info('[ticket:%s] %s 旋转图片.' %
(self.ticket, image_path))
# 旋转90度，并进行文本检测
img_90 = np.rot90(img)
detection_image(self,
img_90,
save_dir,
rotation_angle=90,
file=file,
type='path')
# 旋转270度，并进行文本检测
img_270 = np.rot90(img, 3)
detection_image(self,
img_270,
save_dir,
rotation_angle=270,
file=file,
type='path')
detection_image(self,
img,
save_dir,
rotation_angle=0,
file=file,
type='path')
else:
# 图片文本检测和表格检测
detection_image(self,
img,
save_dir,
rotation_angle=0,
file=file,
type='path')
logger.info('[ticket:%s] detection finished. time cost:%ss' %
(self.ticket, round(time.time() - start_time, 3)))
def result_rearrange(result):
try:
pdf_id_dict = {}
for i, r in result.items():
img_url = r['img_url']
pdf_id = r.get('id', 'pdf_id_default')
page = r.get('page', 0)
if str(page).strip() == '':
page = 0
if pdf_id in pdf_id_dict:
if img_url in pdf_id_dict[pdf_id]:
pdf_id_dict[pdf_id][img_url]['index_list'].append(i)
else:
pdf_id_dict[pdf_id][img_url] = {
'index_list': [i],
"page": page
}
else:
pdf_id_dict[pdf_id] = {
img_url: {
"index_list": [i],
"page": page
}
}
result_new = {}
for pdf_id, values in pdf_id_dict.items():
# 将识别的内容按照pdf页码的顺序进行重排
for im_url, v in sorted(values.items(),
key=lambda x: int(x[1]['page'])):
index_list = v['index_list']
# 从旋转90和270,0中的内容中选一个
if len(index_list) == 3:
index_0 = [
i for i in index_list
if result[str(i)].get('rotation_angle', None) == 0
][0]
index_90 = [
i for i in index_list
if result[str(i)].get('rotation_angle', None) == 90
][0]
index_270 = [
i for i in index_list
if result[str(i)].get('rotation_angle', None) == 270
][0]
_, rotation_angle = \
rotation_content_judg.content_judgment_three(result[str(index_0)]['recoged_result'],
result[str(index_90)]['recoged_result'],
result[str(index_270)]['recoged_result'])
if rotation_angle == 0:
index = index_0
elif rotation_angle == 90:
index = index_90
else:
index = index_270
else:
index = index_list[0]
result_new[str(index)] = result[str(index)]
return result_new
except Exception as e:
logger.info(e)
return result
async def recog_box2word(param):
# model=load_model_with_cuda()
#global model
if not (torch.cuda.is_available() and config.IS_USE_CUDA):
global model
model = model
else:
model = param['model']
self = param['self']
d = param['d']
type = param['type']
use_model = param['use_model']
if use_model:
type = use_model
img_name = d['img_name']
# lock = param['lock']
logger.info('[ticket:%s] %s %s %s model start.' %
(self.ticket, d['img_name'], d['img_url'], type))
start_time = time.time()
img_gray = d['img_gray']
detectied_region = d['detectied_region']
table_bbox = d['table_bbox']
point_sets = d['point_sets']
tmp = {}
# lock.acquire() # 获取锁
recoged_result = eng_recog.box2word(img_gray, detectied_region, model,
type, img_name)
recoged_result = y_axis_deviation_deal(recoged_result)
# lock.release() # 释放锁
tmp['img_url'] = d['img_url']
tmp['rotation_angle'] = d['rotation_angle']
# 文件分类， 只有“其他文件”和“打包文件”入口的文件才需要进行分类，
# 且当“合同”、“发票”，“箱单”，“运单”，“申报要素”，“入库单”等入口有对应文件时，
# 舍弃掉“其他文件”和“打包文件”入口对应的文件
if img_name in ('other_files', 'package_upload'):
# class_time = time.time()
# lock.acquire()
img_name_new = file_class(recoged_result, self.rd, self.img_url_class,
d['img_url'], img_name)
logger.info('[ticket:%s] 文件分类.origin:%s,new:%s' %
(self.ticket, img_name, img_name_new))
# lock.release()
if img_name_new not in ('other_files', 'package_upload'
) and img_name_new in self.source_list:
tmp = {}
elif not recoged_result: # recoged_result的内容为空
tmp = {}
else:
tmp['img_name'] = img_name_new
# result[str(i)] = tmp
else:
tmp['img_name'] = d['img_name']
if recoged_result:
recoged_result = extract_sentence(table_bbox, recoged_result,
d['img_gray'], tmp['img_name'])
tmp['recoged_result'] = recoged_result
tmp['id'] = d.get('id', '')
tmp['page'] = d.get('page', '')
tmp['table_bbox'] = table_bbox
tmp['point_sets'] = point_sets
logger.info('[ticket:%s] %s %s %s model end.time cost:%ss' %
(self.ticket, img_name, d['img_url'], type,
round(time.time() - start_time, 3)))
return tmp if tmp else None
def recognization(self):
# max_workers = (os.cpu_count() - 2) if os.cpu_count() < config.THREADS_NUMBER else config.THREADS_NUMBER
max_workers = (os.cpu_count(
)) if os.cpu_count() < config.THREADS_NUMBER else config.THREADS_NUMBER
logger.info('[ticket:%s] recognization start. max_workers:%s' %
(self.ticket, max_workers))
start_time = time.time()
result_e = {}
result_c = {}
# lock = threading.Lock()
# 多线程
if self.parallel == 1:
# 构建线程池, 线程池默认线程是cpu核数*5， (os.cpu_count() or 1)*5
if torch.cuda.is_available() and config.IS_USE_CUDA: #and set_method:
#mp.set_start_method('spawn')
try:
mp.set_start_method('spawn')
#set_method=False
except RuntimeError:
pass
with ProcessPoolExecutor(max_workers=max_workers) as recog_executor:
new_loop = asyncio.new_event_loop()
asyncio.set_event_loop(new_loop)
event_loop = asyncio.get_event_loop()
try:
ss = time.time()
# all_task_e = []
# all_task_c = []
all_task_params_e = []
all_task_params_c = []
all_task_params_groups = []
##获取中文和英文模型所需参数的列表
for i in range(len(self.detection_list)):
d = self.detection_list[i]
img_name = d['img_name']
use_model = [
i[1] for i in self.rd.chi_tra if img_name == i[0]
]
use_model = use_model[0] if use_model else ''
# 创建英文任务参数列表
if use_model in ('', 'english'):
param_e = {
'self': self,
'd': d,
'i': i,
'type': 'english',
'use_model': use_model
}
# english = recog_executor.submit(recog_box2word, (param_e))
all_task_params_e.append(param_e)
# 创建中文任务参数列表
if use_model in ('', 'chinese', 'chi_tra', 'chi_sim'):
param_c = {
'self': self,
'd': d,
'i': i,
'type': 'chinese',
'use_model': use_model
}
# chinese = recog_executor.submit(recog_box2word, (param_c))
all_task_params_c.append(param_c)
##要将任务分组，让每个进程得到几乎均等的任务，并行的进行计算；此处将参数列表进行分组
# url = 'http://127.0.0.1:5000'
# all_urls = [url for _ in range(100)]
##要加判断两个模型参数都存在的时候才能调用
len_e = len(all_task_params_e)
len_c = len(all_task_params_c)
if len_e > 0:
all_task_params_groups.extend(
chunks(all_task_params_e, recog_executor._max_workers))
if len_c > 0:
all_task_params_groups.extend(
chunks(all_task_params_c, recog_executor._max_workers))
#if torch.cuda.is_available() and config.IS_USE_CUDA:
# if len_e>0:
# all_task_params_groups.extend(chunks(all_task_params_e,recog_executor._max_workers))
# if len_c>0:
# all_task_params_groups.extend(chunks(all_task_params_c,recog_executor._max_workers))
#else:
# if len_e > 0:
# all_task_params_groups.extend(chunks(all_task_params_e, len_e/recog_executor._max_workers))
# if len_c > 0:
# all_task_params_groups.extend(chunks(all_task_params_c, len_c/recog_executor._max_workers))
tasks = [
run(recog_executor, chunked)
for chunked in all_task_params_groups
]
res = event_loop.run_until_complete(asyncio.gather(*tasks))
#将二维列表变成一维列表
res = sum(res, [])
logger.info('{} workers cost time final {} s'.format(
recog_executor._max_workers,
time.time() - ss))
#print('{} workers cost time final'.format(recog_executor._max_workers), time.time() - ss)
# print
# 获取英文任务的执行结果
for index_e, e in enumerate(res[0:len_e]):
result_e[str(index_e)] = e
# 获取中文任务的执行结果
for index_c, c in enumerate(res[len_e:]):
result_c[str(index_c)] = c
except Exception as e:
logger.info('exception err', e, e.errno)
if e.errno != errno.ECONNRESET:
raise
pass
finally:
event_loop.close()
# 单线程
else:
for i in range(len(self.detection_list)):
d = self.detection_list[i]
img_name = d['img_name']
use_model = [i[1] for i in self.rd.chi_tra if img_name == i[0]]
use_model = use_model[0] if use_model else ''
# 英文模型识别
if use_model in ('', 'english'):
param_e = {
'self': self,
'd': d,
'i': i,
'type': 'english',
'use_model': use_model
}
try:
data_e = recog_box2word(param_e)
if data_e:
result_e[str(i)] = data_e
except Exception as e:
logger.info('[ticket:%s] ocr_read_image. %s ' %
(self.ticket, e))
# 中文模型识别
if use_model in ('', 'chinese', 'chi_tra', 'chi_sim'):
param_c = {
'self': self,
'd': d,
'i': i,
'type': 'chinese',
'use_model': use_model
}
try:
data_c = recog_box2word(param_c)
if data_c:
result_c[str(i)] = data_c
except Exception as e:
logger.info('[ticket:%s] ocr_read_image. %s ' %
(self.ticket, e))
if img_class_sencod(self):
result_class_sencod(self, result_e)
result_class_sencod(self, result_c)
if self.data_dir and os.path.exists(self.data_dir):
logger.info('[ticket:%s] 写文件recoged_result.txt' % (self.ticket))
write_content = {'result_e': result_e, 'result_c': result_c}
with open('%s/recoged_result.txt' % self.data_dir,
'w',
encoding='utf8') as f:
f.write(str(write_content) + '\n')
logger.info('[ticket:%s] recognization finished. time cost:%ss' %
(self.ticket, round(time.time() - start_time, 3)))
return result_e, result_c
# if torch.cuda.device_count() > 1:
# model = nn.DataParallel(model)
# # # CPU
# else:
# model.load_state_dict(torch.load(crnn_model_path, map_location='cpu'))
# logger.info('CRNN model loaded.')
#global set_method
#set_method=True
if not (torch.cuda.is_available() and config.IS_USE_CUDA):
crnn_model_path = './models/crnn_Rec_done_99.pth'
alphabet = alphabets.alphabet
nclass = len(alphabet) + 1
# crnn network
model = crnn.CRNN(32, 1, nclass, 256)
model.load_state_dict(torch.load(crnn_model_path, map_location='cpu'))
logger.info('CRNN model loaded.')
else:
logger.info('will CRNN model loaded with gpu.')
def load_model_with_cuda():
crnn_model_path = './models/crnn_Rec_done_99.pth'
alphabet = alphabets.alphabet
nclass = len(alphabet) + 1
# crnn network
model = crnn.CRNN(32, 1, nclass, 256)
#if torch.cuda.is_available() and config.IS_USE_CUDA:
model = model.cuda()
# 导入已经训练好的crnn模型
# # GPU
model.load_state_dict(torch.load(crnn_model_path))
def add_message(uuid):
try:
logger.info('开始进入模型，服务端获取数据')
t1 = time.time()
content = request.get_json(silent=True, force=True)
logger.info(
'from client content:{}'.format(content)) # Do your processing
#response must be a string\tupe\ and so on.
logger.info('将获取的json数据转为dataframe数据框')
# python客户端传来的字符创，网络客户端传来的是dict
logger.info('type content{}'.format(type(content)))
# if isinstance(content,str):
# df_new = pd.read_json(content)
# logger.info('data:{}'.format(df_new.head()))
# logger.info('调用gdp_percent_arima.read_data_new函数')
#
# # data_train,data_test,row_number=gdp_percent_arima.read_data_new(df_new)
# predict_periods=3
# elif isinstance(content,dict):
# # df_new = pd.DataFrame(content)
# #将json数据格式转为模型的输入格式
# logger.info('将客户端传过来的json数据转为算法输入的格式')
# column_name = content['param'][0][0]['historyData'][0]
# logger.info('column_name:{}'.format(column_name))
# data = content['param'][0][0]['historyData'][1:]
# # logger.info('ori_data:{}'.format(np.array(data).shape))
#
# data_train = pd.DataFrame(data=data, columns=column_name)
# logger.info('data_train_shape:{}'.format(data_train))
# # 判断数据是否含有空值和文本内容
# train_res_str_null = judge_null_text.judge_null_str_value(data_train)
# data_train = train_res_str_null['df']
#
#
#
# # name_list = data_train.drop(labels=['date'], axis=1).columns
#
# logger.info('data_train{}'.format(data_train.head()))
#
# test_ori_data = content['param'][1][0]['predictData']
# if test_ori_data != '':
# test_name = content['param'][1][0]['predictData'][0]
# test_data = content['param'][1][0]['predictData'][1:]
# data_test = pd.DataFrame(data=test_data, columns=test_name)
#
# # 判断数据是否含有空值和文本内容
#
# test_res_str_null = judge_null_text.judge_null_str_value(data_test)
#
# data_test = test_res_str_null['df']
#
# #要把训练和测试数据都获取之后再判断是否存在空值，否则训练数据存在空值，那么
# #就不会获取测试数据，那么后面就会显示测试数据变量为定义
# if len(train_res_str_null) > 2:
# raise Exception('存在空值')
#
# if len(test_res_str_null) > 2:
# raise Exception('存在空值')
#
# logger.info('data_test{}'.format(data_test.head()))
# test = ARIMA2.read_data_new(data_test)
#
# # for each_name in test_name:
# # data_test_each = data_test[each_name]
# # logger.info('data_test_each {}'.format(data_test_each.head()))
# else:
# test=False
#
# #本地to_list和tolist都可以用，但是服务器端不可以使用to_list
# logger.info('column_name[0]{}'.format(column_name[0]))
# last_time = data_train[column_name[0]].tolist()[-1]
# logger.info('last_time：{}'.format(last_time))
#
# predict_periods = content['param'][2][0]['predict_periods']
# time_periods_list=range(int(last_time)+1,int(last_time)+int(predict_periods)+1)
# if predict_periods!='':
# try :
# predict_periods=int(predict_periods)
# except Exception as e:
# logger.info('期数转为整型错误{}'.format(e))
# predict_periods = 3
# else:
# logger.info('预测期数未传默认为3期数')
# predict_periods = 3
#
# else:
# logger.info('传入参数类型错误')
#
#
#
# # data_train = data_train['美国']
# # data_test = data_test['美国']
# # name_list=['美国','中国','日本','德国','英国']
#
# all_country_results=[]
# all_country_best_params=[]
# all_relative_error_mean=[]
#
# logger.info('读取训练数据')
# data = ARIMA2.read_data_new(data_train)
#
#
#
# t2 = time.time()
# # 根据参数的顺序来进行选择不同的执行
# for data_order in range(1, len(data)):
# data_subset = data[data_order]
# if test is not False:
# test_data_subset = test[data_order]
# # print('test_data_subset',type(test_data_subset),test_data_subset)
# else:
# test_data_subset = False
#
# # print('dd', data_order, data_subset, '\n\n')
# logger.info('调用ARIMA2模型;{}'.format(data_order))
#
# results = ARIMA2.auto_arima_para_new(data_subset, predict_periods=predict_periods,
# data_test_subset=test_data_subset)
# if results is None:
# logger.info('This problem is unconstrained，无法分析')
# else:
# # print('output_result', results)
# all_country_results.append(results['predict_value'])
# all_country_best_params.append(results['best_params'])
#
# if test is not False:
# all_relative_error_mean.append(results['relative_error_mean'])
# else:
# all_relative_error_mean=['' for each in range(len(data)-1)]
#
#
# logger.info('all_country_results:{}'.format(all_country_results))
# # results.headers['Access-Control-Allow-Origin'] = '*'
# results['success']='0'
# logger.info('len(time_periods_list){}'.format(len(time_periods_list)))
# logger.info('shape{}'.format(np.array(all_country_results).T.shape))
# predict_time_value = np.column_stack((time_periods_list, np.array(all_country_results).T)).tolist()
#
# logger.info('len(all_relative_error_mean){}'.format(len(all_relative_error_mean)))
# logger.info('all_country_best_params shape{}'.format(np.array(all_country_best_params).shape))
#
# all_error_params = np.column_stack((np.array([column_name[1:],all_relative_error_mean]).T, all_country_best_params)).tolist()
#
# logger.info('predict_time_value{}'.format(predict_time_value))
#
# # logger.info('predict_time_value{}'.format(predict_time_value))
# results['predict_value']=predict_time_value
# results['all_error_params']=all_error_params
# logger.info('results:{}'.format(results))
# # print('results',results)
# t3=time.time()
# logger.info('数据处理耗时：{}s'.format(t2 - t1))
# logger.info('模型调用耗时：{}s'.format(t3-t2))
# logger.info('总耗时{}s'.format(t3-t1))
#
#
# # final_results={}
# # final_results['ddd']=results
# # logger.info('final_results:{}'.format(final_results))
return jsonify(content1)
# return jsonify(content)
except Exception as e:
print(e)
app = Flask(__name__, static_url_path='')
# load config from config.py
app.config.from_pyfile('config.py')
url_prefix = app.config.get('url_prefix', '/api/cisdi/ml/economic')
CORS(app, supports_credentials=True)
# CORS(app, resources=r'/*')
@app.route('/')
def index():
return "APIs Server"
app.register_blueprint(bert_qa_blueprint, url_prefix=url_prefix)
# app.register_blueprint(general_arima_blueprint, url_prefix=url_prefix)
# app.register_blueprint(general_corr_blueprint, url_prefix=url_prefix)
# app.register_blueprint(general_regresstion_blueprint, url_prefix=url_prefix)
# app.register_blueprint(special_regresstion_blueprint, url_prefix=url_prefix)
# app.register_blueprint(special_population_blueprint,url_prefix=url_prefix)
if __name__ == '__main__':
host = app.config.get('APP_HOST', 'localhost')
port = app.config.get('APP_PORT', '28095')
logger.info('host:{},port:{}'.format(host, port))
# from werkzeug.contrib.fixers import ProxyFix
# app.wsgi_app = ProxyFix(app.wsgi_app)
# app.run(host=host, port=port, threaded=True, debug=True)
app.run(host=host, debug=False, port=port)
import sys
import os
# print(os.path)
# print(dirname(__file__))
# print(abspath(dirname(__file__)))
#定义搜索路径的优先顺序，序号从0开始，表示最大优先级，sys.path.insert()加入的也是临时搜索路径，程序退出后失效。
# sys.path.insert(0, abspath(dirname(__file__)))
sys.path.insert(0, os.path.dirname(__file__))
# print('nn path',os.path.abspath(os.path.join(os.path.dirname(__file__), '../')))
sys.path.append(os.path.join(os.path.dirname(__file__), '../'))
#__file__主要是解决导入某个模块时，该模块又导入了其他模块这样由于路径导入错误而报错
# print (os.path.abspath(os.path.dirname(__file__)))
from app_logging import logger
logger.info('hello32')
logger.info('哈喽32')
