diff --git a/.gitignore b/.gitignore
index 51f2f2ed8..967c01a0d 100644
--- a/.gitignore
+++ b/.gitignore
@@ -12,3 +12,5 @@ fastdeploy.egg-info
 fastdeploy/version.py
 fastdeploy/LICENSE*
 fastdeploy/ThirdPartyNotices*
+*.so*
+fastdeploy/libs/third_libs
diff --git a/examples/vision/deepcam_yolov5face.cc b/examples/vision/deepcam_yolov5face.cc
new file mode 100644
index 000000000..c6e0083e0
--- /dev/null
+++ b/examples/vision/deepcam_yolov5face.cc
@@ -0,0 +1,53 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision.h"
+
+int main() {
+  namespace vis = fastdeploy::vision;
+
+  std::string model_file = "../resources/models/yolov5s-face.onnx";
+  std::string img_path = "../resources/images/test_face_det.jpg";
+  std::string vis_path =
+      "../resources/outputs/deepcam_yolov5face_vis_result.jpg";
+
+  auto model = vis::deepcam::YOLOv5Face(model_file);
+  if (!model.Initialized()) {
+    std::cerr << "Init Failed! Model: " << model_file << std::endl;
+    return -1;
+  } else {
+    std::cout << "Init Done! Model:" << model_file << std::endl;
+  }
+  model.EnableDebug();
+
+  cv::Mat im = cv::imread(img_path);
+  cv::Mat vis_im = im.clone();
+
+  vis::FaceDetectionResult res;
+  if (!model.Predict(&im, &res, 0.1f, 0.3f)) {
+    std::cerr << "Prediction Failed." << std::endl;
+    return -1;
+  } else {
+    std::cout << "Prediction Done!" << std::endl;
+  }
+
+  // 输出预测框结果
+  std::cout << res.Str() << std::endl;
+
+  // 可视化预测结果
+  vis::Visualize::VisFaceDetection(&vis_im, res, 2, 0.3f);
+  cv::imwrite(vis_path, vis_im);
+  std::cout << "Detect Done! Saved: " << vis_path << std::endl;
+  return 0;
+}
diff --git a/fastdeploy/__init__.py b/fastdeploy/__init__.py
index 68006c1be..948f988b8 100644
--- a/fastdeploy/__init__.py
+++ b/fastdeploy/__init__.py
@@ -32,6 +32,8 @@ def RuntimeOptionStr(runtime_option):
     for attr in attrs:
         if attr.startswith("__"):
             continue
+        if hasattr(getattr(runtime_option, attr), "__call__"):
+            continue
         message += "  {} : {}\t\n".format(attr, getattr(runtime_option, attr))
     message.strip("\n")
     message += ")"
diff --git a/fastdeploy/vision.h b/fastdeploy/vision.h
index 1281df2af..a64e97477 100644
--- a/fastdeploy/vision.h
+++ b/fastdeploy/vision.h
@@ -15,16 +15,17 @@
 
 #include "fastdeploy/core/config.h"
 #ifdef ENABLE_VISION
+#include "fastdeploy/vision/deepcam/yolov5face.h"
 #include "fastdeploy/vision/megvii/yolox.h"
 #include "fastdeploy/vision/meituan/yolov6.h"
 #include "fastdeploy/vision/ppcls/model.h"
 #include "fastdeploy/vision/ppdet/ppyoloe.h"
-#include "fastdeploy/vision/rangilyu/nanodet_plus.h"
 #include "fastdeploy/vision/ppseg/model.h"
+#include "fastdeploy/vision/rangilyu/nanodet_plus.h"
 #include "fastdeploy/vision/ultralytics/yolov5.h"
+#include "fastdeploy/vision/wongkinyiu/scaledyolov4.h"
 #include "fastdeploy/vision/wongkinyiu/yolor.h"
 #include "fastdeploy/vision/wongkinyiu/yolov7.h"
-#include "fastdeploy/vision/wongkinyiu/scaledyolov4.h"
 #endif
 
 #include "fastdeploy/vision/visualize/visualize.h"
diff --git a/fastdeploy/vision/__init__.py b/fastdeploy/vision/__init__.py
index 09be1fa1b..f9c942370 100644
--- a/fastdeploy/vision/__init__.py
+++ b/fastdeploy/vision/__init__.py
@@ -22,4 +22,5 @@ from . import meituan
 from . import megvii
 from . import visualize
 from . import wongkinyiu
+from . import deepcam
 from . import rangilyu
diff --git a/fastdeploy/vision/common/result.cc b/fastdeploy/vision/common/result.cc
index 06a85ea45..0ef077f0c 100644
--- a/fastdeploy/vision/common/result.cc
+++ b/fastdeploy/vision/common/result.cc
@@ -72,6 +72,73 @@ std::string DetectionResult::Str() {
   return out;
 }
 
+FaceDetectionResult::FaceDetectionResult(const FaceDetectionResult& res) {
+  boxes.assign(res.boxes.begin(), res.boxes.end());
+  landmarks.assign(res.landmarks.begin(), res.landmarks.end());
+  scores.assign(res.scores.begin(), res.scores.end());
+  landmarks_per_face = res.landmarks_per_face;
+}
+
+void FaceDetectionResult::Clear() {
+  std::vector<std::array<float, 4>>().swap(boxes);
+  std::vector<float>().swap(scores);
+  std::vector<std::array<float, 2>>().swap(landmarks);
+  landmarks_per_face = 0;
+}
+
+void FaceDetectionResult::Reserve(int size) {
+  boxes.reserve(size);
+  scores.reserve(size);
+  if (landmarks_per_face > 0) {
+    landmarks.reserve(size * landmarks_per_face);
+  }
+}
+
+void FaceDetectionResult::Resize(int size) {
+  boxes.resize(size);
+  scores.resize(size);
+  if (landmarks_per_face > 0) {
+    landmarks.resize(size * landmarks_per_face);
+  }
+}
+
+std::string FaceDetectionResult::Str() {
+  std::string out;
+  // format without landmarks
+  if (landmarks_per_face <= 0) {
+    out = "FaceDetectionResult: [xmin, ymin, xmax, ymax, score]\n";
+    for (size_t i = 0; i < boxes.size(); ++i) {
+      out = out + std::to_string(boxes[i][0]) + "," +
+            std::to_string(boxes[i][1]) + ", " + std::to_string(boxes[i][2]) +
+            ", " + std::to_string(boxes[i][3]) + ", " +
+            std::to_string(scores[i]) + "\n";
+    }
+    return out;
+  }
+  // format with landmarks
+  FDASSERT((landmarks.size() == boxes.size() * landmarks_per_face),
+           "The size of landmarks != boxes.size * landmarks_per_face.");
+  out = "FaceDetectionResult: [xmin, ymin, xmax, ymax, score, (x, y) x " +
+        std::to_string(landmarks_per_face) + "]\n";
+  for (size_t i = 0; i < boxes.size(); ++i) {
+    out = out + std::to_string(boxes[i][0]) + "," +
+          std::to_string(boxes[i][1]) + ", " + std::to_string(boxes[i][2]) +
+          ", " + std::to_string(boxes[i][3]) + ", " +
+          std::to_string(scores[i]) + ", ";
+    for (size_t j = 0; j < landmarks_per_face; ++j) {
+      out = out + "(" +
+            std::to_string(landmarks[i * landmarks_per_face + j][0]) + "," +
+            std::to_string(landmarks[i * landmarks_per_face + j][1]);
+      if (j < landmarks_per_face - 1) {
+        out = out + "), ";
+      } else {
+        out = out + ")\n";
+      }
+    }
+  }
+  return out;
+}
+
 void SegmentationResult::Clear() {
   std::vector<std::vector<int64_t>>().swap(masks);
 }
diff --git a/fastdeploy/vision/common/result.h b/fastdeploy/vision/common/result.h
index 7ff104250..4900d394d 100644
--- a/fastdeploy/vision/common/result.h
+++ b/fastdeploy/vision/common/result.h
@@ -21,7 +21,8 @@ enum FASTDEPLOY_DECL ResultType {
   UNKNOWN_RESULT,
   CLASSIFY,
   DETECTION,
-  SEGMENTATION
+  SEGMENTATION,
+  FACE_DETECTION
 };
 
 struct FASTDEPLOY_DECL BaseResult {
@@ -56,6 +57,31 @@ struct FASTDEPLOY_DECL DetectionResult : public BaseResult {
   std::string Str();
 };
 
+struct FASTDEPLOY_DECL FaceDetectionResult : public BaseResult {
+  // box: xmin, ymin, xmax, ymax
+  std::vector<std::array<float, 4>> boxes;
+  // landmark: x, y, landmarks may empty if the
+  // model don't detect face with landmarks.
+  // Note, one face might have multiple landmarks,
+  // such as 5/19/21/68/98/..., etc.
+  std::vector<std::array<float, 2>> landmarks;
+  std::vector<float> scores;
+  ResultType type = ResultType::FACE_DETECTION;
+  // set landmarks_per_face manually in your post processes.
+  int landmarks_per_face;
+
+  FaceDetectionResult() { landmarks_per_face = 0; }
+  FaceDetectionResult(const FaceDetectionResult& res);
+
+  void Clear();
+
+  void Reserve(int size);
+
+  void Resize(int size);
+
+  std::string Str();
+};
+
 struct FASTDEPLOY_DECL SegmentationResult : public BaseResult {
   // mask
   std::vector<std::vector<int64_t>> masks;
diff --git a/fastdeploy/vision/deepcam/__init__.py b/fastdeploy/vision/deepcam/__init__.py
new file mode 100644
index 000000000..6b1af4328
--- /dev/null
+++ b/fastdeploy/vision/deepcam/__init__.py
@@ -0,0 +1,117 @@
+# Copyright (c) 2022 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 __future__ import absolute_import
+import logging
+from ... import FastDeployModel, Frontend
+from ... import fastdeploy_main as C
+
+
+class YOLOv5Face(FastDeployModel):
+    def __init__(self,
+                 model_file,
+                 params_file="",
+                 runtime_option=None,
+                 model_format=Frontend.ONNX):
+        # 调用基函数进行backend_option的初始化
+        # 初始化后的option保存在self._runtime_option
+        super(YOLOv5Face, self).__init__(runtime_option)
+
+        self._model = C.vision.deepcam.YOLOv5Face(
+            model_file, params_file, self._runtime_option, model_format)
+        # 通过self.initialized判断整个模型的初始化是否成功
+        assert self.initialized, "YOLOv5Face initialize failed."
+
+    def predict(self, input_image, conf_threshold=0.25, nms_iou_threshold=0.5):
+        return self._model.predict(input_image, conf_threshold,
+                                   nms_iou_threshold)
+
+    # 一些跟YOLOv5Face模型有关的属性封装
+    # 多数是预处理相关，可通过修改如model.size = [1280, 1280]改变预处理时resize的大小（前提是模型支持）
+    @property
+    def size(self):
+        return self._model.size
+
+    @property
+    def padding_value(self):
+        return self._model.padding_value
+
+    @property
+    def is_no_pad(self):
+        return self._model.is_no_pad
+
+    @property
+    def is_mini_pad(self):
+        return self._model.is_mini_pad
+
+    @property
+    def is_scale_up(self):
+        return self._model.is_scale_up
+
+    @property
+    def stride(self):
+        return self._model.stride
+
+    @property
+    def landmarks_per_face(self):
+        return self._model.landmarks_per_face
+
+    @size.setter
+    def size(self, wh):
+        assert isinstance(wh, [list, tuple]),\
+            "The value to set `size` must be type of tuple or list."
+        assert len(wh) == 2,\
+            "The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
+            len(wh))
+        self._model.size = wh
+
+    @padding_value.setter
+    def padding_value(self, value):
+        assert isinstance(
+            value,
+            list), "The value to set `padding_value` must be type of list."
+        self._model.padding_value = value
+
+    @is_no_pad.setter
+    def is_no_pad(self, value):
+        assert isinstance(
+            value, bool), "The value to set `is_no_pad` must be type of bool."
+        self._model.is_no_pad = value
+
+    @is_mini_pad.setter
+    def is_mini_pad(self, value):
+        assert isinstance(
+            value,
+            bool), "The value to set `is_mini_pad` must be type of bool."
+        self._model.is_mini_pad = value
+
+    @is_scale_up.setter
+    def is_scale_up(self, value):
+        assert isinstance(
+            value,
+            bool), "The value to set `is_scale_up` must be type of bool."
+        self._model.is_scale_up = value
+
+    @stride.setter
+    def stride(self, value):
+        assert isinstance(
+            value, int), "The value to set `stride` must be type of int."
+        self._model.stride = value
+
+    @landmarks_per_face.setter
+    def landmarks_per_face(self, value):
+        assert isinstance(
+            value,
+            int), "The value to set `landmarks_per_face` must be type of int."
+        self._model.landmarks_per_face = value
diff --git a/fastdeploy/vision/deepcam/deepcam_pybind.cc b/fastdeploy/vision/deepcam/deepcam_pybind.cc
new file mode 100644
index 000000000..3ac741bbc
--- /dev/null
+++ b/fastdeploy/vision/deepcam/deepcam_pybind.cc
@@ -0,0 +1,43 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/pybind/main.h"
+
+namespace fastdeploy {
+void BindDeepCam(pybind11::module& m) {
+  auto deepcam_module =
+      m.def_submodule("deepcam", "https://github.com/deepcam-cn/yolov5-face");
+  pybind11::class_<vision::deepcam::YOLOv5Face, FastDeployModel>(deepcam_module,
+                                                                 "YOLOv5Face")
+      .def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
+      .def("predict",
+           [](vision::deepcam::YOLOv5Face& self, pybind11::array& data,
+              float conf_threshold, float nms_iou_threshold) {
+             auto mat = PyArrayToCvMat(data);
+             vision::FaceDetectionResult res;
+             self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
+             return res;
+           })
+      .def_readwrite("size", &vision::deepcam::YOLOv5Face::size)
+      .def_readwrite("padding_value",
+                     &vision::deepcam::YOLOv5Face::padding_value)
+      .def_readwrite("is_mini_pad", &vision::deepcam::YOLOv5Face::is_mini_pad)
+      .def_readwrite("is_no_pad", &vision::deepcam::YOLOv5Face::is_no_pad)
+      .def_readwrite("is_scale_up", &vision::deepcam::YOLOv5Face::is_scale_up)
+      .def_readwrite("stride", &vision::deepcam::YOLOv5Face::stride)
+      .def_readwrite("landmarks_per_face",
+                     &vision::deepcam::YOLOv5Face::landmarks_per_face);
+}
+
+}  // namespace fastdeploy
diff --git a/fastdeploy/vision/deepcam/yolov5face.cc b/fastdeploy/vision/deepcam/yolov5face.cc
new file mode 100644
index 000000000..5b2c77af9
--- /dev/null
+++ b/fastdeploy/vision/deepcam/yolov5face.cc
@@ -0,0 +1,292 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision/deepcam/yolov5face.h"
+#include "fastdeploy/utils/perf.h"
+#include "fastdeploy/vision/utils/utils.h"
+
+namespace fastdeploy {
+
+namespace vision {
+
+namespace deepcam {
+
+void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
+               bool _auto, bool scale_fill = false, bool scale_up = true,
+               int stride = 32) {
+  float scale =
+      std::min(size[1] * 1.0 / mat->Height(), size[0] * 1.0 / mat->Width());
+  if (!scale_up) {
+    scale = std::min(scale, 1.0f);
+  }
+
+  int resize_h = int(round(mat->Height() * scale));
+  int resize_w = int(round(mat->Width() * scale));
+
+  int pad_w = size[0] - resize_w;
+  int pad_h = size[1] - resize_h;
+  if (_auto) {
+    pad_h = pad_h % stride;
+    pad_w = pad_w % stride;
+  } else if (scale_fill) {
+    pad_h = 0;
+    pad_w = 0;
+    resize_h = size[1];
+    resize_w = size[0];
+  }
+  if (resize_h != mat->Height() || resize_w != mat->Width()) {
+    Resize::Run(mat, resize_w, resize_h);
+  }
+  if (pad_h > 0 || pad_w > 0) {
+    float half_h = pad_h * 1.0 / 2;
+    int top = int(round(half_h - 0.1));
+    int bottom = int(round(half_h + 0.1));
+    float half_w = pad_w * 1.0 / 2;
+    int left = int(round(half_w - 0.1));
+    int right = int(round(half_w + 0.1));
+    Pad::Run(mat, top, bottom, left, right, color);
+  }
+}
+
+YOLOv5Face::YOLOv5Face(const std::string& model_file,
+                       const std::string& params_file,
+                       const RuntimeOption& custom_option,
+                       const Frontend& model_format) {
+  if (model_format == Frontend::ONNX) {
+    valid_cpu_backends = {Backend::ORT};  // 指定可用的CPU后端
+    valid_gpu_backends = {Backend::ORT, Backend::TRT};  // 指定可用的GPU后端
+  } else {
+    valid_cpu_backends = {Backend::PDINFER, Backend::ORT};
+    valid_gpu_backends = {Backend::PDINFER, Backend::ORT, Backend::TRT};
+  }
+  runtime_option = custom_option;
+  runtime_option.model_format = model_format;
+  runtime_option.model_file = model_file;
+  runtime_option.params_file = params_file;
+  initialized = Initialize();
+}
+
+bool YOLOv5Face::Initialize() {
+  // parameters for preprocess
+  size = {640, 640};
+  padding_value = {114.0, 114.0, 114.0};
+  is_mini_pad = false;
+  is_no_pad = false;
+  is_scale_up = false;
+  stride = 32;
+  landmarks_per_face = 5;
+
+  if (!InitRuntime()) {
+    FDERROR << "Failed to initialize fastdeploy backend." << std::endl;
+    return false;
+  }
+  // Check if the input shape is dynamic after Runtime already initialized,
+  // Note that, We need to force is_mini_pad 'false' to keep static
+  // shape after padding (LetterBox) when the is_dynamic_input_ is 'false'.
+  is_dynamic_input_ = false;
+  auto shape = InputInfoOfRuntime(0).shape;
+  for (int i = 0; i < shape.size(); ++i) {
+    // if height or width is dynamic
+    if (i >= 2 && shape[i] <= 0) {
+      is_dynamic_input_ = true;
+      break;
+    }
+  }
+  if (!is_dynamic_input_) {
+    is_mini_pad = false;
+  }
+  return true;
+}
+
+bool YOLOv5Face::Preprocess(
+    Mat* mat, FDTensor* output,
+    std::map<std::string, std::array<float, 2>>* im_info) {
+  // process after image load
+  float ratio = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()),
+                         size[0] * 1.0f / static_cast<float>(mat->Width()));
+  if (ratio != 1.0) {  // always true
+    int interp = cv::INTER_AREA;
+    if (ratio > 1.0) {
+      interp = cv::INTER_LINEAR;
+    }
+    int resize_h = int(round(static_cast<float>(mat->Height()) * ratio));
+    int resize_w = int(round(static_cast<float>(mat->Width()) * ratio));
+    Resize::Run(mat, resize_w, resize_h, -1, -1, interp);
+  }
+  // yolov5face's preprocess steps
+  // 1. letterbox
+  // 2. BGR->RGB
+  // 3. HWC->CHW
+  LetterBox(mat, size, padding_value, is_mini_pad, is_no_pad, is_scale_up,
+            stride);
+  BGR2RGB::Run(mat);
+  // Normalize::Run(mat, std::vector<float>(mat->Channels(), 0.0),
+  //                std::vector<float>(mat->Channels(), 1.0));
+  // Compute `result = mat * alpha + beta` directly by channel
+  std::vector<float> alpha = {1.0f / 255.0f, 1.0f / 255.0f, 1.0f / 255.0f};
+  std::vector<float> beta = {0.0f, 0.0f, 0.0f};
+  Convert::Run(mat, alpha, beta);
+
+  // Record output shape of preprocessed image
+  (*im_info)["output_shape"] = {static_cast<float>(mat->Height()),
+                                static_cast<float>(mat->Width())};
+
+  HWC2CHW::Run(mat);
+  Cast::Run(mat, "float");
+  mat->ShareWithTensor(output);
+  output->shape.insert(output->shape.begin(), 1);  // reshape to n, h, w, c
+  return true;
+}
+
+bool YOLOv5Face::Postprocess(
+    FDTensor& infer_result, FaceDetectionResult* result,
+    const std::map<std::string, std::array<float, 2>>& im_info,
+    float conf_threshold, float nms_iou_threshold) {
+  // infer_result: (1,n,16) 16=4+1+10+1
+  FDASSERT(infer_result.shape[0] == 1, "Only support batch =1 now.");
+  result->Clear();
+  // must be setup landmarks_per_face before reserve
+  result->landmarks_per_face = landmarks_per_face;
+  result->Reserve(infer_result.shape[1]);
+  if (infer_result.dtype != FDDataType::FP32) {
+    FDERROR << "Only support post process with float32 data." << std::endl;
+    return false;
+  }
+  float* data = static_cast<float*>(infer_result.Data());
+  for (size_t i = 0; i < infer_result.shape[1]; ++i) {
+    float* reg_cls_ptr = data + (i * infer_result.shape[2]);
+    float obj_conf = reg_cls_ptr[4];
+    float cls_conf = reg_cls_ptr[15];
+    float confidence = obj_conf * cls_conf;
+    // filter boxes by conf_threshold
+    if (confidence <= conf_threshold) {
+      continue;
+    }
+    float x = reg_cls_ptr[0];
+    float y = reg_cls_ptr[1];
+    float w = reg_cls_ptr[2];
+    float h = reg_cls_ptr[3];
+
+    // convert from [x, y, w, h] to [x1, y1, x2, y2]
+    result->boxes.emplace_back(std::array<float, 4>{
+        (x - w / 2.f), (y - h / 2.f), (x + w / 2.f), (y + h / 2.f)});
+    result->scores.push_back(confidence);
+    // decode landmarks (default 5 landmarks)
+    if (landmarks_per_face > 0) {
+      float* landmarks_ptr = reg_cls_ptr + 5;
+      for (size_t j = 0; j < landmarks_per_face * 2; j += 2) {
+        result->landmarks.emplace_back(
+            std::array<float, 2>{landmarks_ptr[j], landmarks_ptr[j + 1]});
+      }
+    }
+  }
+
+  if (result->boxes.size() == 0) {
+    return true;
+  }
+
+  utils::NMS(result, nms_iou_threshold);
+
+  // scale the boxes to the origin image shape
+  auto iter_out = im_info.find("output_shape");
+  auto iter_ipt = im_info.find("input_shape");
+  FDASSERT(iter_out != im_info.end() && iter_ipt != im_info.end(),
+           "Cannot find input_shape or output_shape from im_info.");
+  float out_h = iter_out->second[0];
+  float out_w = iter_out->second[1];
+  float ipt_h = iter_ipt->second[0];
+  float ipt_w = iter_ipt->second[1];
+  float scale = std::min(out_h / ipt_h, out_w / ipt_w);
+  float pad_h = (out_h - ipt_h * scale) / 2.f;
+  float pad_w = (out_w - ipt_w * scale) / 2.f;
+  if (is_mini_pad) {
+    pad_h = static_cast<float>(static_cast<int>(pad_h) % stride);
+    pad_w = static_cast<float>(static_cast<int>(pad_w) % stride);
+  }
+  // scale and clip box
+  for (size_t i = 0; i < result->boxes.size(); ++i) {
+    result->boxes[i][0] = std::max((result->boxes[i][0] - pad_w) / scale, 0.0f);
+    result->boxes[i][1] = std::max((result->boxes[i][1] - pad_h) / scale, 0.0f);
+    result->boxes[i][2] = std::max((result->boxes[i][2] - pad_w) / scale, 0.0f);
+    result->boxes[i][3] = std::max((result->boxes[i][3] - pad_h) / scale, 0.0f);
+    result->boxes[i][0] = std::min(result->boxes[i][0], ipt_w - 1.0f);
+    result->boxes[i][1] = std::min(result->boxes[i][1], ipt_h - 1.0f);
+    result->boxes[i][2] = std::min(result->boxes[i][2], ipt_w - 1.0f);
+    result->boxes[i][3] = std::min(result->boxes[i][3], ipt_h - 1.0f);
+  }
+  // scale and clip landmarks
+  for (size_t i = 0; i < result->landmarks.size(); ++i) {
+    result->landmarks[i][0] =
+        std::max((result->landmarks[i][0] - pad_w) / scale, 0.0f);
+    result->landmarks[i][1] =
+        std::max((result->landmarks[i][1] - pad_h) / scale, 0.0f);
+    result->landmarks[i][0] = std::min(result->landmarks[i][0], ipt_w - 1.0f);
+    result->landmarks[i][1] = std::min(result->landmarks[i][1], ipt_h - 1.0f);
+  }
+  return true;
+}
+
+bool YOLOv5Face::Predict(cv::Mat* im, FaceDetectionResult* result,
+                         float conf_threshold, float nms_iou_threshold) {
+#ifdef FASTDEPLOY_DEBUG
+  TIMERECORD_START(0)
+#endif
+
+  Mat mat(*im);
+  std::vector<FDTensor> input_tensors(1);
+
+  std::map<std::string, std::array<float, 2>> im_info;
+
+  // Record the shape of image and the shape of preprocessed image
+  im_info["input_shape"] = {static_cast<float>(mat.Height()),
+                            static_cast<float>(mat.Width())};
+  im_info["output_shape"] = {static_cast<float>(mat.Height()),
+                             static_cast<float>(mat.Width())};
+
+  if (!Preprocess(&mat, &input_tensors[0], &im_info)) {
+    FDERROR << "Failed to preprocess input image." << std::endl;
+    return false;
+  }
+
+#ifdef FASTDEPLOY_DEBUG
+  TIMERECORD_END(0, "Preprocess")
+  TIMERECORD_START(1)
+#endif
+
+  input_tensors[0].name = InputInfoOfRuntime(0).name;
+  std::vector<FDTensor> output_tensors;
+  if (!Infer(input_tensors, &output_tensors)) {
+    FDERROR << "Failed to inference." << std::endl;
+    return false;
+  }
+#ifdef FASTDEPLOY_DEBUG
+  TIMERECORD_END(1, "Inference")
+  TIMERECORD_START(2)
+#endif
+
+  if (!Postprocess(output_tensors[0], result, im_info, conf_threshold,
+                   nms_iou_threshold)) {
+    FDERROR << "Failed to post process." << std::endl;
+    return false;
+  }
+
+#ifdef FASTDEPLOY_DEBUG
+  TIMERECORD_END(2, "Postprocess")
+#endif
+  return true;
+}
+
+}  // namespace deepcam
+}  // namespace vision
+}  // namespace fastdeploy
\ No newline at end of file
diff --git a/fastdeploy/vision/deepcam/yolov5face.h b/fastdeploy/vision/deepcam/yolov5face.h
new file mode 100644
index 000000000..74a6f9c69
--- /dev/null
+++ b/fastdeploy/vision/deepcam/yolov5face.h
@@ -0,0 +1,97 @@
+// Copyright (c) 2022 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.
+
+#pragma once
+#include "fastdeploy/fastdeploy_model.h"
+#include "fastdeploy/vision/common/processors/transform.h"
+#include "fastdeploy/vision/common/result.h"
+
+namespace fastdeploy {
+
+namespace vision {
+
+namespace deepcam {
+
+class FASTDEPLOY_DECL YOLOv5Face : public FastDeployModel {
+ public:
+  // 当model_format为ONNX时，无需指定params_file
+  // 当model_format为Paddle时，则需同时指定model_file & params_file
+  YOLOv5Face(const std::string& model_file, const std::string& params_file = "",
+             const RuntimeOption& custom_option = RuntimeOption(),
+             const Frontend& model_format = Frontend::ONNX);
+
+  // 定义模型的名称
+  std::string ModelName() const { return "deepcam-cn/yolov5-face"; }
+
+  // 模型预测接口，即用户调用的接口
+  // im 为用户的输入数据，目前对于CV均定义为cv::Mat
+  // result 为模型预测的输出结构体
+  // conf_threshold 为后处理的参数
+  // nms_iou_threshold 为后处理的参数
+  virtual bool Predict(cv::Mat* im, FaceDetectionResult* result,
+                       float conf_threshold = 0.25,
+                       float nms_iou_threshold = 0.5);
+
+  // 以下为模型在预测时的一些参数，基本是前后处理所需
+  // 用户在创建模型后，可根据模型的要求，以及自己的需求
+  // 对参数进行修改
+  // tuple of (width, height)
+  std::vector<int> size;
+  // padding value, size should be same with Channels
+  std::vector<float> padding_value;
+  // only pad to the minimum rectange which height and width is times of stride
+  bool is_mini_pad;
+  // while is_mini_pad = false and is_no_pad = true, will resize the image to
+  // the set size
+  bool is_no_pad;
+  // if is_scale_up is false, the input image only can be zoom out, the maximum
+  // resize scale cannot exceed 1.0
+  bool is_scale_up;
+  // padding stride, for is_mini_pad
+  int stride;
+  // setup the number of landmarks for per face (if have), default 5 in
+  // official yolov5face note that, the outupt tensor's shape must be:
+  // (1,n,4+1+2*landmarks_per_face+1=box+obj+landmarks+cls)
+  int landmarks_per_face;
+
+ private:
+  // 初始化函数，包括初始化后端，以及其它模型推理需要涉及的操作
+  bool Initialize();
+
+  // 输入图像预处理操作
+  // Mat为FastDeploy定义的数据结构
+  // FDTensor为预处理后的Tensor数据，传给后端进行推理
+  // im_info为预处理过程保存的数据，在后处理中需要用到
+  bool Preprocess(Mat* mat, FDTensor* outputs,
+                  std::map<std::string, std::array<float, 2>>* im_info);
+
+  // 后端推理结果后处理，输出给用户
+  // infer_result 为后端推理后的输出Tensor
+  // result 为模型预测的结果
+  // im_info 为预处理记录的信息，后处理用于还原box
+  // conf_threshold 后处理时过滤box的置信度阈值
+  // nms_iou_threshold 后处理时NMS设定的iou阈值
+  bool Postprocess(FDTensor& infer_result, FaceDetectionResult* result,
+                   const std::map<std::string, std::array<float, 2>>& im_info,
+                   float conf_threshold, float nms_iou_threshold);
+
+  // 查看输入是否为动态维度的 不建议直接使用 不同模型的逻辑可能不一致
+  bool IsDynamicInput() const { return is_dynamic_input_; }
+
+  bool is_dynamic_input_;
+};
+
+}  // namespace deepcam
+}  // namespace vision
+}  // namespace fastdeploy
diff --git a/fastdeploy/vision/utils/nms.cc b/fastdeploy/vision/utils/nms.cc
index d0cd1d59e..900acf84d 100644
--- a/fastdeploy/vision/utils/nms.cc
+++ b/fastdeploy/vision/utils/nms.cc
@@ -66,6 +66,62 @@ void NMS(DetectionResult* result, float iou_threshold) {
   }
 }
 
+void NMS(FaceDetectionResult* result, float iou_threshold) {
+  utils::SortDetectionResult(result);
+
+  std::vector<float> area_of_boxes(result->boxes.size());
+  std::vector<int> suppressed(result->boxes.size(), 0);
+  for (size_t i = 0; i < result->boxes.size(); ++i) {
+    area_of_boxes[i] = (result->boxes[i][2] - result->boxes[i][0]) *
+                       (result->boxes[i][3] - result->boxes[i][1]);
+  }
+
+  for (size_t i = 0; i < result->boxes.size(); ++i) {
+    if (suppressed[i] == 1) {
+      continue;
+    }
+    for (size_t j = i + 1; j < result->boxes.size(); ++j) {
+      if (suppressed[j] == 1) {
+        continue;
+      }
+      float xmin = std::max(result->boxes[i][0], result->boxes[j][0]);
+      float ymin = std::max(result->boxes[i][1], result->boxes[j][1]);
+      float xmax = std::min(result->boxes[i][2], result->boxes[j][2]);
+      float ymax = std::min(result->boxes[i][3], result->boxes[j][3]);
+      float overlap_w = std::max(0.0f, xmax - xmin);
+      float overlap_h = std::max(0.0f, ymax - ymin);
+      float overlap_area = overlap_w * overlap_h;
+      float overlap_ratio =
+          overlap_area / (area_of_boxes[i] + area_of_boxes[j] - overlap_area);
+      if (overlap_ratio > iou_threshold) {
+        suppressed[j] = 1;
+      }
+    }
+  }
+  FaceDetectionResult backup(*result);
+  int landmarks_per_face = result->landmarks_per_face;
+
+  result->Clear();
+  // don't forget to reset the landmarks_per_face
+  // before apply Reserve method.
+  result->landmarks_per_face = landmarks_per_face;
+  result->Reserve(suppressed.size());
+  for (size_t i = 0; i < suppressed.size(); ++i) {
+    if (suppressed[i] == 1) {
+      continue;
+    }
+    result->boxes.emplace_back(backup.boxes[i]);
+    result->scores.push_back(backup.scores[i]);
+    // landmarks (if have)
+    if (result->landmarks_per_face > 0) {
+      for (size_t j = 0; j < result->landmarks_per_face; ++j) {
+        result->landmarks.emplace_back(
+            backup.landmarks[i * result->landmarks_per_face + j]);
+      }
+    }
+  }
+}
+
 }  // namespace utils
 }  // namespace vision
 }  // namespace fastdeploy
diff --git a/fastdeploy/vision/utils/sort_face_det_res.cc b/fastdeploy/vision/utils/sort_face_det_res.cc
new file mode 100644
index 000000000..34150f9ac
--- /dev/null
+++ b/fastdeploy/vision/utils/sort_face_det_res.cc
@@ -0,0 +1,69 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision/utils/utils.h"
+
+namespace fastdeploy {
+namespace vision {
+namespace utils {
+
+void SortDetectionResult(FaceDetectionResult* result) {
+  // sort face detection results with landmarks or not.
+  if (result->boxes.size() == 0) {
+    return;
+  }
+  int landmarks_per_face = result->landmarks_per_face;
+  if (landmarks_per_face > 0) {
+    FDASSERT(
+        (result->landmarks.size() == result->boxes.size() * landmarks_per_face),
+        "The size of landmarks != boxes.size * landmarks_per_face.");
+  }
+
+  // argsort for scores.
+  std::vector<size_t> indices;
+  indices.resize(result->boxes.size());
+  for (size_t i = 0; i < result->boxes.size(); ++i) {
+    indices[i] = i;
+  }
+  std::vector<float>& scores = result->scores;
+  std::sort(indices.begin(), indices.end(),
+            [&scores](size_t a, size_t b) { return scores[a] > scores[b]; });
+
+  // reorder boxes, scores, landmarks (if have).
+  FaceDetectionResult backup(*result);
+  result->Clear();
+  // don't forget to reset the landmarks_per_face
+  // before apply Reserve method.
+  result->landmarks_per_face = landmarks_per_face;
+  result->Reserve(indices.size());
+  if (landmarks_per_face > 0) {
+    for (size_t i = 0; i < indices.size(); ++i) {
+      result->boxes.emplace_back(backup.boxes[indices[i]]);
+      result->scores.push_back(backup.scores[indices[i]]);
+      for (size_t j = 0; j < landmarks_per_face; ++j) {
+        result->landmarks.emplace_back(
+            backup.landmarks[indices[i] * landmarks_per_face + j]);
+      }
+    }
+  } else {
+    for (size_t i = 0; i < indices.size(); ++i) {
+      result->boxes.emplace_back(backup.boxes[indices[i]]);
+      result->scores.push_back(backup.scores[indices[i]]);
+    }
+  }
+}
+
+}  // namespace utils
+}  // namespace vision
+}  // namespace fastdeploy
diff --git a/fastdeploy/vision/utils/utils.h b/fastdeploy/vision/utils/utils.h
index 79ece458c..e95e7e10b 100644
--- a/fastdeploy/vision/utils/utils.h
+++ b/fastdeploy/vision/utils/utils.h
@@ -53,9 +53,13 @@ std::vector<int32_t> TopKIndices(const T* array, int array_size, int topk) {
 
 void NMS(DetectionResult* output, float iou_threshold = 0.5);
 
+void NMS(FaceDetectionResult* result, float iou_threshold = 0.5);
+
 // MergeSort
 void SortDetectionResult(DetectionResult* output);
 
+void SortDetectionResult(FaceDetectionResult* result);
+
 }  // namespace utils
 }  // namespace vision
 }  // namespace fastdeploy
diff --git a/fastdeploy/vision/vision_pybind.cc b/fastdeploy/vision/vision_pybind.cc
index 42fcebff4..3b426ebd8 100644
--- a/fastdeploy/vision/vision_pybind.cc
+++ b/fastdeploy/vision/vision_pybind.cc
@@ -23,6 +23,7 @@ void BindPPSeg(pybind11::module& m);
 void BindUltralytics(pybind11::module& m);
 void BindMeituan(pybind11::module& m);
 void BindMegvii(pybind11::module& m);
+void BindDeepCam(pybind11::module& m);
 void BindRangiLyu(pybind11::module& m);
 #ifdef ENABLE_VISION_VISUALIZE
 void BindVisualize(pybind11::module& m);
@@ -44,6 +45,15 @@ void BindVision(pybind11::module& m) {
       .def("__repr__", &vision::DetectionResult::Str)
       .def("__str__", &vision::DetectionResult::Str);
 
+  pybind11::class_<vision::FaceDetectionResult>(m, "FaceDetectionResult")
+      .def(pybind11::init())
+      .def_readwrite("boxes", &vision::FaceDetectionResult::boxes)
+      .def_readwrite("scores", &vision::FaceDetectionResult::scores)
+      .def_readwrite("landmarks", &vision::FaceDetectionResult::landmarks)
+      .def_readwrite("landmarks_per_face",
+                     &vision::FaceDetectionResult::landmarks_per_face)
+      .def("__repr__", &vision::FaceDetectionResult::Str)
+      .def("__str__", &vision::FaceDetectionResult::Str);
   pybind11::class_<vision::SegmentationResult>(m, "SegmentationResult")
       .def(pybind11::init())
       .def_readwrite("masks", &vision::SegmentationResult::masks)
@@ -57,6 +67,7 @@ void BindVision(pybind11::module& m) {
   BindWongkinyiu(m);
   BindMeituan(m);
   BindMegvii(m);
+  BindDeepCam(m);
   BindRangiLyu(m);
 #ifdef ENABLE_VISION_VISUALIZE
   BindVisualize(m);
diff --git a/fastdeploy/vision/visualize/__init__.py b/fastdeploy/vision/visualize/__init__.py
index 7d1bcc892..ea836b68e 100644
--- a/fastdeploy/vision/visualize/__init__.py
+++ b/fastdeploy/vision/visualize/__init__.py
@@ -21,6 +21,11 @@ def vis_detection(im_data, det_result, line_size=1, font_size=0.5):
     C.vision.Visualize.vis_detection(im_data, det_result, line_size, font_size)
 
 
+def vis_face_detection(im_data, face_det_result, line_size=1, font_size=0.5):
+    C.vision.Visualize.vis_face_detection(im_data, face_det_result, line_size,
+                                          font_size)
+
+
 def vis_segmentation(im_data, seg_result, vis_im_data, num_classes=1000):
     C.vision.Visualize.vis_segmentation(im_data, seg_result, vis_im_data,
                                         num_classes)
diff --git a/fastdeploy/vision/visualize/face_detection.cc b/fastdeploy/vision/visualize/face_detection.cc
new file mode 100644
index 000000000..8a95a1ad7
--- /dev/null
+++ b/fastdeploy/vision/visualize/face_detection.cc
@@ -0,0 +1,81 @@
+// Copyright (c) 2022 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.
+
+#ifdef ENABLE_VISION_VISUALIZE
+
+#include "fastdeploy/vision/visualize/visualize.h"
+#include "opencv2/imgproc/imgproc.hpp"
+
+namespace fastdeploy {
+
+namespace vision {
+
+// Default only support visualize num_classes <= 1000
+// If need to visualize num_classes > 1000
+// Please call Visualize::GetColorMap(num_classes) first
+void Visualize::VisFaceDetection(cv::Mat* im, const FaceDetectionResult& result,
+                                 int line_size, float font_size) {
+  auto color_map = GetColorMap();
+  int h = im->rows;
+  int w = im->cols;
+
+  bool vis_landmarks = false;
+  if ((result.landmarks_per_face > 0) &&
+      (result.boxes.size() * result.landmarks_per_face ==
+       result.landmarks.size())) {
+    vis_landmarks = true;
+  }
+  for (size_t i = 0; i < result.boxes.size(); ++i) {
+    cv::Rect rect(result.boxes[i][0], result.boxes[i][1],
+                  result.boxes[i][2] - result.boxes[i][0],
+                  result.boxes[i][3] - result.boxes[i][1]);
+    int color_id = i % 333;
+    int c0 = color_map[3 * color_id + 0];
+    int c1 = color_map[3 * color_id + 1];
+    int c2 = color_map[3 * color_id + 2];
+    cv::Scalar rect_color = cv::Scalar(c0, c1, c2);
+    std::string text = std::to_string(result.scores[i]);
+    if (text.size() > 4) {
+      text = text.substr(0, 4);
+    }
+    int font = cv::FONT_HERSHEY_SIMPLEX;
+    cv::Size text_size = cv::getTextSize(text, font, font_size, 1, nullptr);
+    cv::Point origin;
+    origin.x = rect.x;
+    origin.y = rect.y;
+    cv::Rect text_background =
+        cv::Rect(result.boxes[i][0], result.boxes[i][1] - text_size.height,
+                 text_size.width, text_size.height);
+    cv::rectangle(*im, rect, rect_color, line_size);
+    cv::putText(*im, text, origin, font, font_size, cv::Scalar(255, 255, 255),
+                1);
+    // vis landmarks (if have)
+    if (vis_landmarks) {
+      cv::Scalar landmark_color = rect_color;
+      for (size_t j = 0; j < result.landmarks_per_face; ++j) {
+        cv::Point landmark;
+        landmark.x = static_cast<int>(
+            result.landmarks[i * result.landmarks_per_face + j][0]);
+        landmark.y = static_cast<int>(
+            result.landmarks[i * result.landmarks_per_face + j][1]);
+        cv::circle(*im, landmark, line_size, landmark_color, -1);
+      }
+    }
+  }
+}
+
+}  // namespace vision
+}  // namespace fastdeploy
+
+#endif
\ No newline at end of file
diff --git a/fastdeploy/vision/visualize/visualize.h b/fastdeploy/vision/visualize/visualize.h
index 1eb212c2b..f0fab5ee9 100644
--- a/fastdeploy/vision/visualize/visualize.h
+++ b/fastdeploy/vision/visualize/visualize.h
@@ -27,6 +27,8 @@ class FASTDEPLOY_DECL Visualize {
   static const std::vector<int>& GetColorMap(int num_classes = 1000);
   static void VisDetection(cv::Mat* im, const DetectionResult& result,
                            int line_size = 2, float font_size = 0.5f);
+  static void VisFaceDetection(cv::Mat* im, const FaceDetectionResult& result,
+                               int line_size = 2, float font_size = 0.5f);
   static void VisSegmentation(const cv::Mat& im,
                               const SegmentationResult& result,
                               cv::Mat* vis_img, const int& num_classes = 1000);
diff --git a/fastdeploy/vision/visualize/visualize_pybind.cc b/fastdeploy/vision/visualize/visualize_pybind.cc
index 5d5eb2388..4e12f55c8 100644
--- a/fastdeploy/vision/visualize/visualize_pybind.cc
+++ b/fastdeploy/vision/visualize/visualize_pybind.cc
@@ -25,6 +25,14 @@ void BindVisualize(pybind11::module& m) {
                     vision::Visualize::VisDetection(&im, result, line_size,
                                                     font_size);
                   })
+      .def_static(
+          "vis_face_detection",
+          [](pybind11::array& im_data, vision::FaceDetectionResult& result,
+             int line_size, float font_size) {
+            auto im = PyArrayToCvMat(im_data);
+            vision::Visualize::VisFaceDetection(&im, result, line_size,
+                                                font_size);
+          })
       .def_static("vis_segmentation", [](pybind11::array& im_data,
                                          vision::SegmentationResult& result,
                                          pybind11::array& vis_im_data,
diff --git a/fastdeploy/vision/wongkinyiu/scaledyolov4.cc b/fastdeploy/vision/wongkinyiu/scaledyolov4.cc
index 240e3b7ba..7321fc01b 100644
--- a/fastdeploy/vision/wongkinyiu/scaledyolov4.cc
+++ b/fastdeploy/vision/wongkinyiu/scaledyolov4.cc
@@ -21,8 +21,8 @@ namespace vision {
 namespace wongkinyiu {
 
 void ScaledYOLOv4::LetterBox(Mat* mat, const std::vector<int>& size,
-                      const std::vector<float>& color, bool _auto,
-                      bool scale_fill, bool scale_up, int stride) {
+                             const std::vector<float>& color, bool _auto,
+                             bool scale_fill, bool scale_up, int stride) {
   float scale =
       std::min(size[1] * 1.0 / mat->Height(), size[0] * 1.0 / mat->Width());
   if (!scale_up) {
@@ -57,8 +57,10 @@ void ScaledYOLOv4::LetterBox(Mat* mat, const std::vector<int>& size,
   }
 }
 
-ScaledYOLOv4::ScaledYOLOv4(const std::string& model_file, const std::string& params_file,
-             const RuntimeOption& custom_option, const Frontend& model_format) {
+ScaledYOLOv4::ScaledYOLOv4(const std::string& model_file,
+                           const std::string& params_file,
+                           const RuntimeOption& custom_option,
+                           const Frontend& model_format) {
   if (model_format == Frontend::ONNX) {
     valid_cpu_backends = {Backend::ORT};  // 指定可用的CPU后端
     valid_gpu_backends = {Backend::ORT, Backend::TRT};  // 指定可用的GPU后端
@@ -90,8 +92,9 @@ bool ScaledYOLOv4::Initialize() {
   return true;
 }
 
-bool ScaledYOLOv4::Preprocess(Mat* mat, FDTensor* output,
-                       std::map<std::string, std::array<float, 2>>* im_info) {
+bool ScaledYOLOv4::Preprocess(
+    Mat* mat, FDTensor* output,
+    std::map<std::string, std::array<float, 2>>* im_info) {
   // process after image load
   float ratio = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()),
                          size[0] * 1.0f / static_cast<float>(mat->Width()));
@@ -109,7 +112,7 @@ bool ScaledYOLOv4::Preprocess(Mat* mat, FDTensor* output,
   // 2. BGR->RGB
   // 3. HWC->CHW
   ScaledYOLOv4::LetterBox(mat, size, padding_value, is_mini_pad, is_no_pad,
-                   is_scale_up, stride);
+                          is_scale_up, stride);
   BGR2RGB::Run(mat);
   // Normalize::Run(mat, std::vector<float>(mat->Channels(), 0.0),
   //                std::vector<float>(mat->Channels(), 1.0));
@@ -176,7 +179,7 @@ bool ScaledYOLOv4::Postprocess(
   float pad_h = (out_h - ipt_h * scale) / 2.0f;
   float pad_w = (out_w - ipt_w * scale) / 2.0f;
   if (is_mini_pad) {
-    // 和 LetterBox中_auto=true的处理逻辑对应 
+    // 和 LetterBox中_auto=true的处理逻辑对应
     pad_h = static_cast<float>(static_cast<int>(pad_h) % stride);
     pad_w = static_cast<float>(static_cast<int>(pad_w) % stride);
   }
@@ -199,8 +202,8 @@ bool ScaledYOLOv4::Postprocess(
   return true;
 }
 
-bool ScaledYOLOv4::Predict(cv::Mat* im, DetectionResult* result, float conf_threshold,
-                    float nms_iou_threshold) {
+bool ScaledYOLOv4::Predict(cv::Mat* im, DetectionResult* result,
+                           float conf_threshold, float nms_iou_threshold) {
 #ifdef FASTDEPLOY_DEBUG
   TIMERECORD_START(0)
 #endif
diff --git a/fastdeploy/vision/wongkinyiu/scaledyolov4.h b/fastdeploy/vision/wongkinyiu/scaledyolov4.h
index c85fc8a04..788b57474 100644
--- a/fastdeploy/vision/wongkinyiu/scaledyolov4.h
+++ b/fastdeploy/vision/wongkinyiu/scaledyolov4.h
@@ -25,9 +25,10 @@ class FASTDEPLOY_DECL ScaledYOLOv4 : public FastDeployModel {
  public:
   // 当model_format为ONNX时，无需指定params_file
   // 当model_format为Paddle时，则需同时指定model_file & params_file
-  ScaledYOLOv4(const std::string& model_file, const std::string& params_file = "",
-        const RuntimeOption& custom_option = RuntimeOption(),
-        const Frontend& model_format = Frontend::ONNX);
+  ScaledYOLOv4(const std::string& model_file,
+               const std::string& params_file = "",
+               const RuntimeOption& custom_option = RuntimeOption(),
+               const Frontend& model_format = Frontend::ONNX);
 
   // 定义模型的名称
   virtual std::string ModelName() const { return "WongKinYiu/ScaledYOLOv4"; }
diff --git a/model_zoo/vision/yolov5face/README.md b/model_zoo/vision/yolov5face/README.md
new file mode 100644
index 000000000..e1713e67d
--- /dev/null
+++ b/model_zoo/vision/yolov5face/README.md
@@ -0,0 +1,78 @@
+# YOLOv5Face部署示例
+
+当前支持模型版本为：[YOLOv5Face CommitID:4fd1ead](https://github.com/deepcam-cn/yolov5-face/commit/4fd1ead)
+
+本文档说明如何进行[YOLOv5Face](https://github.com/deepcam-cn/yolov5-face)的快速部署推理。本目录结构如下
+
+```
+.
+├── cpp                     # C++ 代码目录
+│   ├── CMakeLists.txt      # C++ 代码编译CMakeLists文件
+│   ├── README.md           # C++ 代码编译部署文档
+│   └── yolov5face.cc       # C++ 示例代码
+├── api.md                  # API 说明文档
+├── README.md               # YOLOv5Face 部署文档
+└── yolov5face.py           # Python示例代码
+```
+
+## 获取ONNX文件
+
+访问[YOLOv5Face](https://github.com/deepcam-cn/yolov5-face)官方github库，按照指引下载安装，下载`yolov5s-face.pt` 模型，利用 `export.py` 得到`onnx`格式文件。
+
+* 下载yolov5face模型文件
+  ```
+  Link: https://pan.baidu.com/s/1fyzLxZYx7Ja1_PCIWRhxbw Link: eq0q  
+  https://drive.google.com/file/d/1zxaHeLDyID9YU4-hqK7KNepXIwbTkRIO/view?usp=sharing
+  ```
+
+* 导出onnx格式文件
+  ```bash
+  PYTHONPATH=. python export.py --weights weights/yolov5s-face.pt --img_size 640 640 --batch_size 1  
+  ```
+* onnx模型简化(可选)
+  ```bash
+  onnxsim yolov5s-face.onnx yolov5s-face.onnx
+  ```
+* 移动onnx文件到model_zoo/yolov5face的目录
+  ```bash
+  cp PATH/TO/yolov5s-face.onnx PATH/TO/model_zoo/vision/yolov5face/
+  ```
+
+
+
+## 准备测试图片
+准备一张包含人脸的测试图片，命名为test.jpg，并拷贝到可执行文件所在的目录
+
+## 安装FastDeploy
+
+使用如下命令安装FastDeploy，注意到此处安装的是`vision-cpu`，也可根据需求安装`vision-gpu`
+```bash
+# 安装fastdeploy-python工具
+pip install fastdeploy-python
+
+# 安装vision-cpu模块
+fastdeploy install vision-cpu
+```
+
+## Python部署
+
+执行如下代码即会自动下载YOLOv5Face模型和测试图片
+```bash
+python yolov5face.py
+```
+
+执行完成后会将可视化结果保存在本地`vis_result.jpg`，同时输出检测结果如下
+```
+FaceDetectionResult: [xmin, ymin, xmax, ymax, score, (x, y) x 5]
+749.575256,375.122162, 775.008850, 407.858215, 0.851824, (756.933838,388.423157), (767.810974,387.932922), (762.617065,394.212341), (758.053101,399.073639), (767.370300,398.769470)
+897.833862,380.372864, 924.725281, 409.566803, 0.847505, (903.757202,390.221741), (914.575867,389.495911), (908.998901,395.983307), (905.803223,400.871429), (914.674438,400.268066)
+281.558197,367.739349, 305.474701, 397.860535, 0.840915, (287.018768,379.771088), (297.285004,378.755280), (292.057831,385.207367), (289.110962,390.010437), (297.535339,389.412048)
+132.922104,368.507263, 159.098541, 402.777283, 0.840232, (140.632492,382.361633), (151.900864,380.966156), (146.869186,388.505066), (141.930420,393.724670), (151.734604,392.808197)
+699.379700,306.743256, 723.219421, 336.533295, 0.840228, (705.688843,319.133301), (715.784668,318.449524), (711.107300,324.416016), (707.236633,328.671936), (716.088623,328.151794)
+# ...
+```
+
+## 其它文档
+
+- [C++部署](./cpp/README.md)
+- [YOLOv5Face API文档](./api.md)
diff --git a/model_zoo/vision/yolov5face/api.md b/model_zoo/vision/yolov5face/api.md
new file mode 100644
index 000000000..384ef23d3
--- /dev/null
+++ b/model_zoo/vision/yolov5face/api.md
@@ -0,0 +1,71 @@
+# YOLOv5Face API说明
+
+## Python API
+
+### YOLOv5Face类
+```
+fastdeploy.vision.deepcam.YOLOv5Face(model_file, params_file=None, runtime_option=None, model_format=fd.Frontend.ONNX)
+```
+YOLOv5Face模型加载和初始化，当model_format为`fd.Frontend.ONNX`时，只需提供model_file，如`yolov5s-face.onnx`；当model_format为`fd.Frontend.PADDLE`时，则需同时提供model_file和params_file。
+
+**参数**
+
+> * **model_file**(str): 模型文件路径
+> * **params_file**(str): 参数文件路径
+> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
+> * **model_format**(Frontend): 模型格式
+
+#### predict函数
+> ```
+> YOLOv5Face.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
+> ```
+> 模型预测结口，输入图像直接输出检测结果。
+>
+> **参数**
+>
+> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
+> > * **conf_threshold**(float): 检测框置信度过滤阈值
+> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
+
+示例代码参考[yolov5face.py](./yolov5face.py)
+
+
+## C++ API
+
+### YOLOv5Face类
+```
+fastdeploy::vision::deepcam::YOLOv5Face(
+        const string& model_file,
+        const string& params_file = "",
+        const RuntimeOption& runtime_option = RuntimeOption(),
+        const Frontend& model_format = Frontend::ONNX)
+```
+YOLOv5Face模型加载和初始化，当model_format为`Frontend::ONNX`时，只需提供model_file，如`yolov5s-face.onnx`；当model_format为`Frontend::PADDLE`时，则需同时提供model_file和params_file。
+
+**参数**
+
+> * **model_file**(str): 模型文件路径
+> * **params_file**(str): 参数文件路径
+> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
+> * **model_format**(Frontend): 模型格式
+
+#### Predict函数
+> ```
+> YOLOv5Face::Predict(cv::Mat* im, DetectionResult* result,
+>                     float conf_threshold = 0.25,
+>                     float nms_iou_threshold = 0.5)
+> ```
+> 模型预测接口，输入图像直接输出检测结果。
+>
+> **参数**
+>
+> > * **im**: 输入图像，注意需为HWC，BGR格式
+> > * **result**: 检测结果，包括检测框，各个框的置信度
+> > * **conf_threshold**: 检测框置信度过滤阈值
+> > * **nms_iou_threshold**: NMS处理过程中iou阈值
+
+示例代码参考[cpp/yolov5face.cc](cpp/yolov5face.cc)
+
+## 其它API使用
+
+- [模型部署RuntimeOption配置](../../../docs/api/runtime_option.md)
diff --git a/model_zoo/vision/yolov5face/cpp/CMakeLists.txt b/model_zoo/vision/yolov5face/cpp/CMakeLists.txt
new file mode 100644
index 000000000..23878ac2c
--- /dev/null
+++ b/model_zoo/vision/yolov5face/cpp/CMakeLists.txt
@@ -0,0 +1,17 @@
+PROJECT(yolov5face_demo C CXX)
+CMAKE_MINIMUM_REQUIRED (VERSION 3.16)
+
+# 在低版本ABI环境中，通过如下代码进行兼容性编译
+# add_definitions(-D_GLIBCXX_USE_CXX11_ABI=0)
+
+# 指定下载解压后的fastdeploy库路径
+set(FASTDEPLOY_INSTALL_DIR ${PROJECT_SOURCE_DIR}/fastdeploy-linux-x64-0.3.0/)
+
+include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
+
+# 添加FastDeploy依赖头文件
+include_directories(${FASTDEPLOY_INCS})
+
+add_executable(yolov5face_demo ${PROJECT_SOURCE_DIR}/yolov5face.cc)
+# 添加FastDeploy库依赖
+target_link_libraries(yolov5face_demo ${FASTDEPLOY_LIBS})
diff --git a/model_zoo/vision/yolov5face/cpp/README.md b/model_zoo/vision/yolov5face/cpp/README.md
new file mode 100644
index 000000000..4f5788458
--- /dev/null
+++ b/model_zoo/vision/yolov5face/cpp/README.md
@@ -0,0 +1,60 @@
+# 编译YOLOv5Face示例
+
+当前支持模型版本为：[YOLOv5Face CommitID:4fd1ead](https://github.com/deepcam-cn/yolov5-face/commit/4fd1ead)
+
+## 下载和解压预测库
+```bash
+wget https://bj.bcebos.com/paddle2onnx/fastdeploy/fastdeploy-linux-x64-0.0.3.tgz
+tar xvf fastdeploy-linux-x64-0.0.3.tgz
+```
+
+## 编译示例代码
+```bash
+mkdir build & cd build
+cmake ..
+make -j
+```
+
+## 获取ONNX文件
+
+访问[YOLOv5Face](https://github.com/deepcam-cn/yolov5-face)官方github库，按照指引下载安装，下载`yolov5s-face.pt` 模型，利用 `export.py` 得到`onnx`格式文件。
+
+* 下载yolov5face模型文件
+  ```
+  Link: https://pan.baidu.com/s/1fyzLxZYx7Ja1_PCIWRhxbw Link: eq0q  
+  https://drive.google.com/file/d/1zxaHeLDyID9YU4-hqK7KNepXIwbTkRIO/view?usp=sharing
+  ```
+
+* 导出onnx格式文件
+  ```bash
+  PYTHONPATH=. python export.py --weights weights/yolov5s-face.pt --img_size 640 640 --batch_size 1  
+  ```
+* onnx模型简化(可选)
+  ```bash
+  onnxsim yolov5s-face.onnx yolov5s-face.onnx
+  ```
+* 移动onnx文件到可执行文件的目录
+  ```bash
+  cp PATH/TO/yolov5s-face.onnx PATH/TO/model_zoo/vision/yolov5face/cpp/build
+  ```
+
+
+
+## 准备测试图片
+准备一张包含人脸的测试图片，命名为test.jpg，并拷贝到可执行文件所在的目录
+
+## 执行
+```bash
+./yolov5face_demo
+```
+
+执行完后可视化的结果保存在本地`vis_result.jpg`，同时会将检测框输出在终端，如下所示
+```
+aceDetectionResult: [xmin, ymin, xmax, ymax, score, (x, y) x 5]
+749.575256,375.122162, 775.008850, 407.858215, 0.851824, (756.933838,388.423157), (767.810974,387.932922), (762.617065,394.212341), (758.053101,399.073639), (767.370300,398.769470)
+897.833862,380.372864, 924.725281, 409.566803, 0.847505, (903.757202,390.221741), (914.575867,389.495911), (908.998901,395.983307), (905.803223,400.871429), (914.674438,400.268066)
+281.558197,367.739349, 305.474701, 397.860535, 0.840915, (287.018768,379.771088), (297.285004,378.755280), (292.057831,385.207367), (289.110962,390.010437), (297.535339,389.412048)
+132.922104,368.507263, 159.098541, 402.777283, 0.840232, (140.632492,382.361633), (151.900864,380.966156), (146.869186,388.505066), (141.930420,393.724670), (151.734604,392.808197)
+699.379700,306.743256, 723.219421, 336.533295, 0.840228, (705.688843,319.133301), (715.784668,318.449524), (711.107300,324.416016), (707.236633,328.671936), (716.088623,328.151794)
+# ...
+```
diff --git a/model_zoo/vision/yolov5face/cpp/yolov5face.cc b/model_zoo/vision/yolov5face/cpp/yolov5face.cc
new file mode 100644
index 000000000..baa0bb7c0
--- /dev/null
+++ b/model_zoo/vision/yolov5face/cpp/yolov5face.cc
@@ -0,0 +1,40 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision.h"
+
+int main() {
+  namespace vis = fastdeploy::vision;
+  auto model = vis::deepcam::YOLOv5Face("yolov5s-face.onnx");
+  if (!model.Initialized()) {
+    std::cerr << "Init Failed." << std::endl;
+    return -1;
+  }
+  cv::Mat im = cv::imread("test.jpg");
+  cv::Mat vis_im = im.clone();
+
+  vis::FaceDetectionResult res;
+  if (!model.Predict(&im, &res, 0.1f, 0.3f)) {
+    std::cerr << "Prediction Failed." << std::endl;
+    return -1;
+  }
+
+  // 输出预测框结果
+  std::cout << res.Str() << std::endl;
+
+  // 可视化预测结果
+  vis::Visualize::VisFaceDetection(&vis_im, res, 2, 0.3f);
+  cv::imwrite("vis_result.jpg", vis_im);
+  return 0;
+}
diff --git a/model_zoo/vision/yolov5face/yolov5face.py b/model_zoo/vision/yolov5face/yolov5face.py
new file mode 100644
index 000000000..ff7ab1b77
--- /dev/null
+++ b/model_zoo/vision/yolov5face/yolov5face.py
@@ -0,0 +1,17 @@
+import fastdeploy as fd
+import cv2
+
+# 加载模型
+model = fd.vision.deepcam.YOLOv5Face("yolov5s-face.onnx")
+
+# 预测图片
+im = cv2.imread("test.jpg")
+result = model.predict(im, conf_threshold=0.1, nms_iou_threshold=0.3)
+
+# 可视化结果
+fd.vision.visualize.vis_face_detection(im, result)
+cv2.imwrite("vis_result.jpg", im)
+
+# 输出预测结果
+print(result)
+print(model.runtime_option)