diff --git a/python/examples/blazeface/README.md b/python/examples/blazeface/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..f569841ce4a3ae69b1ff16041f7fb7d4617177f7
--- /dev/null
+++ b/python/examples/blazeface/README.md
@@ -0,0 +1,23 @@
+# Blazeface 
+
+## Get Model
+```
+python -m paddle_serving_app.package --get_model blazeface
+tar -xzvf blazeface.tar.gz
+```
+
+## RPC Service
+
+### Start Service
+
+```
+python -m paddle_serving_server.serve --model serving_server --port 9494
+```
+
+### Client Prediction
+
+```
+python test_client.py serving_client/serving_client_conf.prototxt test.jpg
+```
+
+the result is in `output` folder, including a json file and image file with bounding boxes.
diff --git a/python/examples/blazeface/test_client.py b/python/examples/blazeface/test_client.py
new file mode 100644
index 0000000000000000000000000000000000000000..27eb185ea90ce72641cef44d9066c46945ad2629
--- /dev/null
+++ b/python/examples/blazeface/test_client.py
@@ -0,0 +1,38 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from paddle_serving_client import Client
+from paddle_serving_app.reader import *
+import sys
+import numpy as np
+
+preprocess = Sequential([
+    File2Image(),
+    Normalize([104, 117, 123], [127.502231, 127.502231, 127.502231], False)
+])
+
+postprocess = BlazeFacePostprocess("label_list.txt", "output")
+client = Client()
+
+client.load_client_config(sys.argv[1])
+client.connect(['127.0.0.1:9494'])
+
+im_0 = preprocess(sys.argv[2])
+tmp = Transpose((2, 0, 1))
+im = tmp(im_0)
+fetch_map = client.predict(
+    feed={"image": im}, fetch=["detection_output_0.tmp_0"])
+fetch_map["image"] = sys.argv[2]
+fetch_map["im_shape"] = im_0.shape
+postprocess(fetch_map)
diff --git a/python/paddle_serving_app/models/model_list.py b/python/paddle_serving_app/models/model_list.py
index 79b3f91bd6584d17ddbc4124584cf40bd586b965..c5e78c25e554ef0438c90f489a3f726b4f87ceaa 100644
--- a/python/paddle_serving_app/models/model_list.py
+++ b/python/paddle_serving_app/models/model_list.py
@@ -24,7 +24,8 @@ class ServingModels(object):
             "SentimentAnalysis"] = ["senta_bilstm", "senta_bow", "senta_cnn"]
         self.model_dict["SemanticRepresentation"] = ["ernie"]
         self.model_dict["ChineseWordSegmentation"] = ["lac"]
-        self.model_dict["ObjectDetection"] = ["faster_rcnn", "yolov4"]
+        self.model_dict[
+            "ObjectDetection"] = ["faster_rcnn", "yolov4", "blazeface"]
         self.model_dict["ImageSegmentation"] = [
             "unet", "deeplabv3", "deeplabv3+cityscapes"
         ]
diff --git a/python/paddle_serving_app/reader/functional.py b/python/paddle_serving_app/reader/functional.py
index 4240641dd99fceb278ff60a5ba1dbb5275e534aa..7bab279c7f1aa71a2d55a8cb7b12bcb38607eb70 100644
--- a/python/paddle_serving_app/reader/functional.py
+++ b/python/paddle_serving_app/reader/functional.py
@@ -29,6 +29,7 @@ def normalize(img, mean, std, channel_first):
     else:
         img_mean = np.array(mean).reshape((1, 1, 3))
         img_std = np.array(std).reshape((1, 1, 3))
+    img = np.array(img).astype("float32")
     img -= img_mean
     img /= img_std
     return img
diff --git a/python/paddle_serving_app/reader/image_reader.py b/python/paddle_serving_app/reader/image_reader.py
index a44ca5de84da2bafce9b4cea37fb88095debabc6..096f46549af137cb04a87e26a3b28c8d42e33daa 100644
--- a/python/paddle_serving_app/reader/image_reader.py
+++ b/python/paddle_serving_app/reader/image_reader.py
@@ -440,6 +440,30 @@ class RCNNPostprocess(object):
             self.label_file, self.output_dir)
 
 
+class BlazeFacePostprocess(RCNNPostprocess):
+    def clip_bbox(self, bbox, im_size=None):
+        h = 1. if im_size is None else im_size[0]
+        w = 1. if im_size is None else im_size[1]
+        xmin = max(min(bbox[0], w), 0.)
+        ymin = max(min(bbox[1], h), 0.)
+        xmax = max(min(bbox[2], w), 0.)
+        ymax = max(min(bbox[3], h), 0.)
+        return xmin, ymin, xmax, ymax
+
+    def _get_bbox_result(self, fetch_map, fetch_name, clsid2catid):
+        result = {}
+        is_bbox_normalized = True  #for blaze face, set true here
+        output = fetch_map[fetch_name]
+        lod = [fetch_map[fetch_name + '.lod']]
+        lengths = self._offset_to_lengths(lod)
+        np_data = np.array(output)
+        result['bbox'] = (np_data, lengths)
+        result['im_id'] = np.array([[0]])
+        result["im_shape"] = np.array(fetch_map["im_shape"]).astype(np.int32)
+        bbox_results = self._bbox2out([result], clsid2catid, is_bbox_normalized)
+        return bbox_results
+
+
 class Sequential(object):
     """
     Args: