[Model] Add PIPNet and FaceLandmark1000 Support

examples/vision/facealign/README.md

@@ -5,3 +5,5 @@ FastDeploy目前支持如下人脸对齐(关键点检测）模型部署
 | 模型 | 说明 | 模型格式 | 版本 |
 | :--- | :--- | :------- | :--- |
 | [Hsintao/pfld_106_face_landmarks](./pfld) | PFLD 系列模型 | ONNX | [CommitID:e150195](https://github.com/Hsintao/pfld_106_face_landmarks/commit/e150195) |
+| [Single430/FaceLandmark1000](./face_landmark_1000) | FaceLandmark1000 系列模型 | ONNX | [CommitID:1a951b6](https://github.com/Single430/FaceLandmark1000/tree/1a951b6) |
+| [jhb86253817/PIPNet](./pipnet) | PIPNet 系列模型 | ONNX | [CommitID:b9eab58](https://github.com/jhb86253817/PIPNet/tree/b9eab58) |

examples/vision/facealign/face_landmark_1000/README.md

@@ -0,0 +1,25 @@
+# FaceLandmark 模型部署
+
+## 模型版本说明
+
+- [FaceLandmark1000](https://github.com/Single430/FaceLandmark1000/tree/1a951b6)
+
+## 支持模型列表
+
+目前FastDeploy支持如下模型的部署
+
+- [FaceLandmark1000 模型](https://github.com/Single430/FaceLandmark1000)
+
+## 下载预训练模型
+
+为了方便开发者的测试，下面提供了FaceLandmark导出的各系列模型，开发者可直接下载使用。
+
+| 模型                                                               | 参数大小    | 精度    | 备注 |
+|:---------------------------------------------------------------- |:----- |:----- | :------ |
+| [FaceLandmark1000](https://bj.bcebos.com/paddlehub/fastdeploy/FaceLandmark1000.onnx) | 2.1M | - |
+
+
+## 详细部署文档
+
+- [Python部署](python)
+- [C++部署](cpp)

examples/vision/facealign/face_landmark_1000/cpp/CMakeLists.txt

@@ -0,0 +1,14 @@
+PROJECT(infer_demo C CXX)
+CMAKE_MINIMUM_REQUIRED (VERSION 3.10)
+
+# 指定下载解压后的fastdeploy库路径
+option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
+include(${FASTDEPLOY_INSTALL_DIR}/utils/gflags.cmake)
+include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
+
+# 添加FastDeploy依赖头文件
+include_directories(${FASTDEPLOY_INCS})
+
+add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
+# 添加FastDeploy库依赖
+target_link_libraries(infer_demo ${FASTDEPLOY_LIBS} gflags pthread)

examples/vision/facealign/face_landmark_1000/cpp/README.md

@@ -0,0 +1,84 @@
+# FaceLandmark1000 C++部署示例
+
+本目录下提供`infer.cc`快速完成FaceLandmark1000在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
+
+在部署前，需确认以下两个步骤
+
+- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md)  
+- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md)
+
+以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试，保证 FastDeploy 版本0.6.0以上(x.x.x >= 0.6.0)支持FaceLandmark1000模型
+
+```bash
+mkdir build
+cd build
+wget https://bj.bcebos.com/fastdeploy/release/cpp/fastdeploy-linux-x64-x.x.x.tgz
+tar xvf fastdeploy-linux-x64-x.x.x.tgz
+cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-x.x.x
+make -j
+
+#下载官方转换好的 FaceLandmark1000 模型文件和测试图片
+wget https://bj.bcebos.com/paddlehub/fastdeploy/FaceLandmark1000.onnx
+wget https://bj.bcebos.com/paddlehub/fastdeploy/facealign_input.png
+
+# CPU推理
+./infer_demo --model FaceLandmark1000.onnx --image facealign_input.png --device cpu
+# GPU推理
+./infer_demo --model FaceLandmark1000.onnx --image facealign_input.png --device gpu
+# GPU上TensorRT推理
+./infer_demo --model FaceLandmark1000.onnx --image facealign_input.png --device gpu --backend trt
+```
+
+运行完成可视化结果如下图所示
+
+<div width="500">
+<img width="470" height="384" float="left" src="https://user-images.githubusercontent.com/67993288/200761309-90c096e2-c2f3-4140-8012-32ed84e5f389.jpg">
+</div>
+
+以上命令只适用于Linux或MacOS, Windows下SDK的使用方式请参考:  
+- [如何在Windows中使用FastDeploy C++ SDK](../../../../../docs/cn/faq/use_sdk_on_windows.md)
+
+## FaceLandmark1000 C++接口
+
+### FaceLandmark1000 类
+
+```c++
+fastdeploy::vision::facealign::FaceLandmark1000(
+        const string& model_file,
+        const string& params_file = "",
+        const RuntimeOption& runtime_option = RuntimeOption(),
+        const ModelFormat& model_format = ModelFormat::ONNX)
+```
+
+FaceLandmark1000模型加载和初始化，其中model_file为导出的ONNX模型格式。
+
+**参数**
+
+> * **model_file**(str): 模型文件路径
+> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
+> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
+> * **model_format**(ModelFormat): 模型格式，默认为ONNX格式
+
+#### Predict函数
+
+> ```c++
+> FaceLandmark1000::Predict(cv::Mat* im, FaceAlignmentResult* result)
+> ```
+>
+> 模型预测接口，输入图像直接输出landmarks结果。
+>
+> **参数**
+>
+> > * **im**: 输入图像，注意需为HWC，BGR格式
+> > * **result**: landmarks结果, FaceAlignmentResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
+
+### 类成员变量
+
+用户可按照自己的实际需求，修改下列预处理参数，从而影响最终的推理和部署效果
+
+> > * **size**(vector&lt;int&gt;): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[128, 128]
+
+- [模型介绍](../../)
+- [Python部署](../python)
+- [视觉模型预测结果](../../../../../docs/api/vision_results/)
+- [如何切换模型推理后端引擎](../../../../../docs/cn/faq/how_to_change_backend.md)

examples/vision/facealign/face_landmark_1000/cpp/infer.cc

@@ -0,0 +1,110 @@
+// 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"
+#include "gflags/gflags.h"
+
+DEFINE_string(model, "", "Directory of the inference model.");
+DEFINE_string(image, "", "Path of the image file.");
+DEFINE_string(device, "cpu",
+              "Type of inference device, support 'cpu' or 'gpu'.");
+DEFINE_string(backend, "default",
+              "The inference runtime backend, support: ['default', 'ort', "
+              "'paddle', 'ov', 'trt', 'paddle_trt']");
+DEFINE_bool(use_fp16, false, "Whether to use FP16 mode, only support 'trt' and 'paddle_trt' backend");
+
+void PrintUsage() {
+  std::cout << "Usage: infer_demo --model model_path --image img_path --device [cpu|gpu] --backend "
+               "[default|ort|paddle|ov|trt|paddle_trt] "
+               "--use_fp16 false"
+            << std::endl;
+  std::cout << "Default value of device: cpu" << std::endl;
+  std::cout << "Default value of backend: default" << std::endl;
+  std::cout << "Default value of use_fp16: false" << std::endl;
+}
+
+bool CreateRuntimeOption(fastdeploy::RuntimeOption* option) {
+  if (FLAGS_device == "gpu") {
+    option->UseGpu();
+    if (FLAGS_backend == "ort") {
+      option->UseOrtBackend();
+    } else if (FLAGS_backend == "paddle") {
+      option->UsePaddleBackend();
+    } else if (FLAGS_backend == "trt" || 
+               FLAGS_backend == "paddle_trt") {
+      option->UseTrtBackend();
+      option->SetTrtInputShape("input", {1, 3, 128, 128});
+      if (FLAGS_backend == "paddle_trt") {
+        option->EnablePaddleToTrt();
+      }
+      if (FLAGS_use_fp16) {
+        option->EnableTrtFP16();
+      }
+    } else if (FLAGS_backend == "default") {
+      return true;
+    } else {
+      std::cout << "While inference with GPU, only support default/ort/paddle/trt/paddle_trt now, " << FLAGS_backend << " is not supported." << std::endl;
+      return false;
+    }
+  } else if (FLAGS_device == "cpu") {
+    if (FLAGS_backend == "ort") {
+      option->UseOrtBackend();
+    } else if (FLAGS_backend == "ov") {
+      option->UseOpenVINOBackend();
+    } else if (FLAGS_backend == "paddle") {
+      option->UsePaddleBackend();
+    } else if (FLAGS_backend == "default") {
+      return true;
+    } else {
+      std::cout << "While inference with CPU, only support default/ort/ov/paddle now, " << FLAGS_backend << " is not supported." << std::endl;
+      return false;
+    }
+  } else {
+    std::cerr << "Only support device CPU/GPU now, "  << FLAGS_device << " is not supported." << std::endl;
+    return false;
+  }
+
+  return true;
+}
+
+int main(int argc, char* argv[]) {
+  google::ParseCommandLineFlags(&argc, &argv, true);
+  auto option = fastdeploy::RuntimeOption();
+  if (!CreateRuntimeOption(&option)) {
+    PrintUsage();
+    return -1;
+  }
+
+  auto model = fastdeploy::vision::facealign::FaceLandmark1000(FLAGS_model, "", option);
+  if (!model.Initialized()) {
+    std::cerr << "Failed to initialize." << std::endl;
+    return -1;
+  }
+
+  auto im = cv::imread(FLAGS_image);
+  auto im_bak = im.clone();
+
+  fastdeploy::vision::FaceAlignmentResult res;
+  if (!model.Predict(&im, &res)) {
+    std::cerr << "Failed to predict." << std::endl;
+    return -1;
+  }
+  std::cout << res.Str() << std::endl;
+
+  auto vis_im = fastdeploy::vision::VisFaceAlignment(im_bak, res);
+  cv::imwrite("vis_result.jpg", vis_im);
+  std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
+
+  return 0;
+}

examples/vision/facealign/face_landmark_1000/python/README.md

@@ -0,0 +1,71 @@
+# FaceLandmark1000 Python部署示例
+
+在部署前，需确认以下两个步骤
+
+- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md)  
+- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md)
+
+本目录下提供`infer.py`快速完成FaceLandmark1000在CPU/GPU，以及GPU上通过TensorRT加速部署的示例，保证 FastDeploy 版本 >= 0.6.0 支持FaceLandmark1000模型。执行如下脚本即可完成
+
+```bash
+#下载部署示例代码
+git clone https://github.com/PaddlePaddle/FastDeploy.git
+cd FastDeploy/examples/vision/facealign/facelandmark1000/python
+
+# 下载FaceLandmark1000模型文件和测试图片
+## 原版ONNX模型
+wget https://bj.bcebos.com/paddlehub/fastdeploy/FaceLandmark1000.onnx
+wget https://bj.bcebos.com/paddlehub/fastdeploy/facealign_input.png
+
+# CPU推理
+python infer.py --model FaceLandmark1000.onnx --image facealign_input.png --device cpu
+# GPU推理
+python infer.py --model FaceLandmark1000.onnx --image facealign_input.png --device gpu
+# TRT推理
+python infer.py --model FaceLandmark1000.onnx --image facealign_input.png --device gpu --backend trt
+```
+
+运行完成可视化结果如下图所示
+
+<div width="500">
+<img width="470" height="384" float="left" src="https://user-images.githubusercontent.com/67993288/200761309-90c096e2-c2f3-4140-8012-32ed84e5f389.jpg">
+</div>
+
+## FaceLandmark1000 Python接口
+
+```python
+fd.vision.facealign.FaceLandmark1000(model_file, params_file=None, runtime_option=None, model_format=ModelFormat.ONNX)
+```
+
+FaceLandmark1000模型加载和初始化，其中model_file为导出的ONNX模型格式
+
+**参数**
+
+> * **model_file**(str): 模型文件路径
+> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
+> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
+> * **model_format**(ModelFormat): 模型格式，默认为ONNX
+
+### predict函数
+
+> ```python
+> FaceLandmark1000.predict(input_image)
+> ```
+>
+> 模型预测结口，输入图像直接输出landmarks坐标结果。
+>
+> **参数**
+>
+> > * **input_image**(np.ndarray): 输入数据，注意需为HWC，BGR格式
+
+> **返回**
+>
+> > 返回`fastdeploy.vision.FaceAlignmentResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
+
+
+## 其它文档
+
+- [FaceLandmark1000 模型介绍](..)
+- [FaceLandmark1000 C++部署](../cpp)
+- [模型预测结果说明](../../../../../docs/api/vision_results/)
+- [如何切换模型推理后端引擎](../../../../../docs/cn/faq/how_to_change_backend.md)