diff --git a/paddlehub/commands/serving.py b/paddlehub/commands/serving.py
index 81cd75316880d63ce41ac9628e5d52030b19e811..8dced0f13091d4380b274b4a2052cc238e59498c 100644
--- a/paddlehub/commands/serving.py
+++ b/paddlehub/commands/serving.py
@@ -19,6 +19,7 @@ from __future__ import print_function
 
 import argparse
 import os
+import platform
 import socket
 import json
 import paddlehub as hub
@@ -39,9 +40,9 @@ class ServingCommand(BaseCommand):
             usage='%(prog)s',
             add_help=True)
         self.parser.add_argument("command")
+        self.parser.add_argument("sub_command")
         self.sub_parse = self.parser.add_mutually_exclusive_group(
             required=False)
-        self.sub_parse.add_argument("--start", action="store_true")
         self.parser.add_argument(
             "--use_gpu", action="store_true", default=False)
         self.parser.add_argument(
@@ -51,7 +52,7 @@ class ServingCommand(BaseCommand):
         self.parser.add_argument("--port", "-p", nargs="+", default=[8866])
 
     @staticmethod
-    def port_is_open(ip, port):
+    def is_port_occupied(ip, port):
         s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
         try:
             s.connect((ip, int(port)))
@@ -93,19 +94,28 @@ class ServingCommand(BaseCommand):
 
     @staticmethod
     def start_serving(args):
-        if args.use_multiprocess is True:
-            from paddlehub.serving import app
-        else:
-            from paddlehub.serving import app_single as app
         config_file = args.config
         if config_file is not None:
             config_file = config_file[0]
             if os.path.exists(config_file):
                 with open(config_file, "r") as fp:
                     configs = json.load(fp)
+                    use_multiprocess = configs.get("use_multiprocess", False)
+                    if use_multiprocess is True:
+                        if platform.system() == "Windows":
+                            print(
+                                "Warning: Windows cannot use multiprocess working "
+                                "mode, Hub-Serving will switch to single process mode"
+                            )
+                            from paddlehub.serving import app_single as app
+                        else:
+                            from paddlehub.serving import app
+                    else:
+                        from paddlehub.serving import app_single as app
                     use_gpu = configs.get("use_gpu", False)
                     port = configs.get("port", 8866)
-                    if ServingCommand.port_is_open("127.0.0.1", port) is True:
+                    if ServingCommand.is_port_occupied("127.0.0.1",
+                                                       port) is True:
                         print("Port %s is occupied, please change it." % (port))
                         return False
                     configs = configs.get("modules_info")
@@ -120,11 +130,21 @@ class ServingCommand(BaseCommand):
             else:
                 print("config_file ", config_file, "not exists.")
         else:
+            if args.use_multiprocess is True:
+                if platform.system() == "Windows":
+                    print(
+                        "Warning: Windows cannot use multiprocess working "
+                        "mode, Hub-Serving will switch to single process mode")
+                    from paddlehub.serving import app_single as app
+                else:
+                    from paddlehub.serving import app
+            else:
+                from paddlehub.serving import app_single as app
             module = args.modules
             if module is not None:
                 use_gpu = args.use_gpu
                 port = args.port[0]
-                if ServingCommand.port_is_open("127.0.0.1", port) is True:
+                if ServingCommand.is_port_occupied("127.0.0.1", port) is True:
                     print("Port %s is occupied, please change it." % (port))
                     return False
                 module_info = ServingCommand.preinstall_modules(module)
@@ -142,9 +162,10 @@ class ServingCommand(BaseCommand):
     def show_help():
         str = "serving <option>\n"
         str += "\tManage PaddleHub-Serving.\n"
-        str += "option:\n"
-        str += "--start\n"
+        str += "sub command:\n"
+        str += "start\n"
         str += "\tStart PaddleHub-Serving if specifies this parameter.\n"
+        str += "option:\n"
         str += "--modules/-m [module1==version, module2==version...]\n"
         str += "\tPre-install modules via this parameter list.\n"
         str += "--port/-p XXXX\n"
@@ -163,7 +184,7 @@ class ServingCommand(BaseCommand):
             print("Please refer to the instructions below.")
             ServingCommand.show_help()
             return False
-        if args.start is True:
+        if args.sub_command == "start":
             ServingCommand.start_serving(args)
         else:
             ServingCommand.show_help()
diff --git a/paddlehub/serving/app_single.py b/paddlehub/serving/app_single.py
index 66d34333da38a242c37d9c71458d3f52c7a4c610..794e71b0223e8123fea993f4f812ee7a75aef1aa 100644
--- a/paddlehub/serving/app_single.py
+++ b/paddlehub/serving/app_single.py
@@ -22,6 +22,48 @@ import os
 import base64
 import logging
 
+nlp_module_method = {
+    "lac": "predict_lexical_analysis",
+    "simnet_bow": "predict_sentiment_analysis",
+    "lm_lstm": "predict_pretrained_model",
+    "senta_lstm": "predict_pretrained_model",
+    "senta_gru": "predict_pretrained_model",
+    "senta_cnn": "predict_pretrained_model",
+    "senta_bow": "predict_pretrained_model",
+    "senta_bilstm": "predict_pretrained_model",
+    "emotion_detection_textcnn": "predict_pretrained_model"
+}
+cv_module_method = {
+    "vgg19_imagenet": "predict_classification",
+    "vgg16_imagenet": "predict_classification",
+    "vgg13_imagenet": "predict_classification",
+    "vgg11_imagenet": "predict_classification",
+    "shufflenet_v2_imagenet": "predict_classification",
+    "se_resnext50_32x4d_imagenet": "predict_classification",
+    "se_resnext101_32x4d_imagenet": "predict_classification",
+    "resnet_v2_50_imagenet": "predict_classification",
+    "resnet_v2_34_imagenet": "predict_classification",
+    "resnet_v2_18_imagenet": "predict_classification",
+    "resnet_v2_152_imagenet": "predict_classification",
+    "resnet_v2_101_imagenet": "predict_classification",
+    "pnasnet_imagenet": "predict_classification",
+    "nasnet_imagenet": "predict_classification",
+    "mobilenet_v2_imagenet": "predict_classification",
+    "googlenet_imagenet": "predict_classification",
+    "alexnet_imagenet": "predict_classification",
+    "yolov3_coco2017": "predict_object_detection",
+    "ultra_light_fast_generic_face_detector_1mb_640":
+    "predict_object_detection",
+    "ultra_light_fast_generic_face_detector_1mb_320":
+    "predict_object_detection",
+    "ssd_mobilenet_v1_pascal": "predict_object_detection",
+    "pyramidbox_face_detection": "predict_object_detection",
+    "faster_rcnn_coco2017": "predict_object_detection",
+    "cyclegan_cityscapes": "predict_gan",
+    "deeplabv3p_xception65_humanseg": "predict_semantic_segmentation",
+    "ace2p": "predict_semantic_segmentation"
+}
+
 
 def predict_sentiment_analysis(module, input_text, extra=None):
     global use_gpu
@@ -221,7 +263,7 @@ def create_app():
         global use_gpu
         img_base64 = request.form.getlist("image")
         file_name_list = []
-        if img_base64 != "":
+        if img_base64 != []:
             for item in img_base64:
                 ext = item.split(";")[0].split("/")[-1]
                 if ext not in ["jpeg", "jpg", "png"]:
@@ -241,8 +283,12 @@ def create_app():
                 item.save(file_name)
                 file_name_list.append(file_name)
         module = ImageModelService.get_module(module_name)
-        module_type = module.type.split("/")[-1].replace("-", "_").lower()
-        predict_func = eval("predict_" + module_type)
+        predict_func_name = cv_module_method.get(module_name, "")
+        if predict_func_name != "":
+            predict_func = eval(predict_func_name)
+        else:
+            module_type = module.type.split("/")[-1].replace("-", "_").lower()
+            predict_func = eval("predict_" + module_type)
         results = predict_func(module, file_name_list)
         r = {"results": str(results)}
         return r
@@ -259,8 +305,12 @@ def create_app():
             data = request.form.getlist("text")
         file = request.files.getlist("user_dict")
         module = TextModelService.get_module(module_name)
-        module_type = module.type.split("/")[-1].replace("-", "_").lower()
-        predict_func = eval("predict_" + module_type)
+        predict_func_name = nlp_module_method.get(module_name, "")
+        if predict_func_name != "":
+            predict_func = eval(predict_func_name)
+        else:
+            module_type = module.type.split("/")[-1].replace("-", "_").lower()
+            predict_func = eval("predict_" + module_type)
         file_list = []
         for item in file:
             file_path = req_id + "_" + item.filename
diff --git a/paddlehub/serving/templates/serving_config.json b/paddlehub/serving/templates/serving_config.json
index d26dd6b6fbdfecf477ffb217d8a061ba0d51e4f7..d7254da918806a673f24f874367c4b94525540d1 100644
--- a/paddlehub/serving/templates/serving_config.json
+++ b/paddlehub/serving/templates/serving_config.json
@@ -26,5 +26,6 @@
   }
 ],
   "use_gpu": false,
-  "port": 8866
+  "port": 8866,
+  "use_multiprocess": false
 }