### Setup Linux Environment for GPUPixel

Source: https://gpupixel.pixpark.net/zh/guide/build

Run this script to install necessary dependencies for compiling GPUPixel on Debian or Ubuntu systems.

```bash
./script/setup_env_linux.sh
```

--------------------------------

### Compile GPUPixel for Windows

Source: https://gpupixel.pixpark.net/zh/guide/build

This batch file compiles the GPUPixel library for Windows. Ensure Visual Studio 2017+ and CMake 3.10+ are installed. Only supports x86_64 Release builds.

```bash
.\script\build_windows.bat
```

--------------------------------

### Adjust Beauty Face Filter Parameters

Source: https://gpupixel.pixpark.net/zh/call/beauty_effects

Provides examples for adjusting the skin smoothing and whitening intensity of the BeautyFaceFilter. Values are typically normalized between 0.0 and 1.0.

```cpp
// 设置磨皮程度 (0.0-1.0)
beauty_face_filter_->SetBlurAlpha(value/10);

// 设置美白程度 (0.0-1.0)
beauty_face_filter_->SetWhite(value/20);
```

--------------------------------

### Build GPUPixel for Windows and Linux

Source: https://gpupixel.pixpark.net/zh/guide/demo

Run the platform-specific build scripts from the project root directory to compile the demo.

```bash
.\script\build_windows.bat
```

```bash
./script/build_linux.sh
```

--------------------------------

### Initialize Custom Input Source

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Implement the static `Create` method and constructor for your custom input source. Ensure proper initialization by calling `Init()` and returning a valid shared pointer.

```cpp
std::shared_ptr<YourCustomInput> YourCustomInput::Create() {
  auto input = std::shared_ptr<YourCustomInput>(new YourCustomInput());
  if (input->Init()) {
    return input;
  }
  return nullptr;
}
```

--------------------------------

### Integrate Custom Input Source

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Create an instance of your custom input source using `YourCustomInput::Create()`. Add sinks (filters) to the input source and set up further processing or output.

```cpp
auto input = YourCustomInput::Create();
if (input) {
  auto filter = BeautyFaceFilter::Create();
  input->AddSink(filter);
  // Set more filters or output
}
```

--------------------------------

### Compile GPUPixel for macOS

Source: https://gpupixel.pixpark.net/zh/guide/build

Run this script from the project root to compile the GPUPixel library for macOS. The compiled output will be in the 'output' directory.

```bash
./script/build_macos.sh
```

--------------------------------

### Compile GPUPixel for iOS

Source: https://gpupixel.pixpark.net/zh/guide/build

Execute this script from the project root to compile the GPUPixel library for iOS. Output is located in the 'output' directory.

```bash
./script/build_ios.sh
```

--------------------------------

### 创建并链接 GPUPixel 滤镜链

Source: https://gpupixel.pixpark.net/zh/call/basic_filters

通过实例化源、滤镜和目标组件，并使用 AddSink 方法将它们串联起来以构建处理链路。

```cpp
// 声明组件
std::shared_ptr<SourceRawData> source_raw_input_;
std::shared_ptr<BeautyFaceFilter> beauty_face_filter_;
std::shared_ptr<SinkRawData> target_raw_output_;

// 初始化并链接组件
// 创建源
source_raw_input_ = SourceRawData::Create();

// 创建滤镜
beauty_face_filter_ = BeautyFaceFilter::Create();

// 创建目标
target_raw_output_ = SinkRawData::Create();

// 链接滤镜链
source_raw_input_->AddSink(beauty_face_filter_)
                    ->AddSink(target_raw_output_);
```

--------------------------------

### Compile GPUPixel for Linux

Source: https://gpupixel.pixpark.net/zh/guide/build

Execute this script from the project root to compile the GPUPixel library on Linux (Debian/Ubuntu). Output is located in the 'output' directory.

```bash
./script/build_linux.sh
```

--------------------------------

### Create and Chain Face Beauty Filters

Source: https://gpupixel.pixpark.net/zh/call/beauty_effects

Demonstrates how to create and link multiple beauty filters in a chain for complex effects. Ensure face detection is enabled if using face-specific filters.

```cpp
   // 创建源
   gpuPixelRawInput = SourceRawData::Create();

   // 创建目标
   target_raw_output_ = SinkRawData::Create();

   // 创建滤镜
   lipstick_filter_ = LipstickFilter::Create();
   blusher_filter_ = BlusherFilter::Create();
   face_reshape_filter_ = FaceReshapeFilter::Create();
   beauty_face_filter_ = BeautyFaceFilter::Create();

   // 创建人脸检测器（需启用 GPUPIXEL_ENABLE_FACE_DETECTOR）
   face_detector_ = FaceDetector::Create();

   // 链接滤镜链
   gpuPixelRawInput->AddSink(lipstick_filter_)
                  ->AddSink(blusher_filter_)
                  ->AddSink(face_reshape_filter_)
                  ->AddSink(beauty_face_filter_)
                  ->AddSink(target_raw_output_);
```

--------------------------------

### Add GPUPixel Framework to Xcode Project

Source: https://gpupixel.pixpark.net/zh/guide/integrated

Steps to add the gpupixel.framework to your iOS or MacOS project's build phases and search paths.

```bash
├── gpupixel.framework
```

```bash
├── gpupixel.framework
```

--------------------------------

### Initialize Custom Filter with Shader and Parameters in C++

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

The Init() method initializes the filter with a fragment shader string and registers custom properties. Ensure the shader initialization is successful and parameters are registered with their setters.

```cpp
bool CustomFilter::Init() {
  // 使用着色器字符串初始化
  if (!InitWithFragmentShaderString(kCustomFragmentShader)) {
    return false;
  }

  // 注册参数
  _parameter = 1.0f;
  RegisterProperty(
      "parameter",
      _parameter,
      "参数描述",
      [this](float& value) { setParameter(value); });

  return true;
}
```

--------------------------------

### Define Custom Input Source Class

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Inherit from the `Source` class to create a custom input source. Implement a static `Create` method and private members as needed.

```cpp
class YourCustomInput : public Source {
 public:
  static std::shared_ptr<YourCustomInput> Create();
  // Your custom methods
 private:
  // Your private members
};
```

--------------------------------

### Compile GPUPixel for Android

Source: https://gpupixel.pixpark.net/zh/guide/build

Use this script from the project root to compile the GPUPixel library for Android. The output AAR file is located in 'src/android/java/gpupixel/build/outputs/aar'.

```bash
./script/build_android.sh
```

--------------------------------

### Process Frame with Face Detection and Filter Chain

Source: https://gpupixel.pixpark.net/zh/call/beauty_effects

Shows how to detect face landmarks using a detector and pass them to relevant filters before processing the frame. This is crucial for real-time facial adjustments.

```cpp
   // 获取当前帧图像数据（宽、高、stride 等）
   const uint8_t* buffer = ...;  // 例如从相机或其它输入获取
   int width = ..., height = ..., stride = ...;

   std::vector<float> landmarks = face_detector_->Detect(
       buffer, width, height, stride,
       GPUPIXEL_MODE_FMT_PICTURE, GPUPIXEL_FRAME_TYPE_RGBA);

   if (!landmarks.empty()) {
      lipstick_filter_->SetFaceLandmarks(landmarks);
      blusher_filter_->SetFaceLandmarks(landmarks);
      face_reshape_filter_->SetFaceLandmarks(landmarks);
   }

   // 再执行滤镜链：使用 SourceRawData 时调用 ProcessData 送入本帧并触发渲染；
   // 若使用 SourceImage，则改为先取 buffer = source->GetRgbaImageBuffer() 再 source->Render()
   gpuPixelRawInput->ProcessData(buffer, width, height, stride,
                                GPUPIXEL_FRAME_TYPE_RGBA);
```

--------------------------------

### Implement Custom Filter Factory Method in C++

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

Implement a static factory method to create and initialize a custom filter instance. It ensures proper initialization within the GPUPixel context.

```cpp
std::shared_ptr<CustomFilter> CustomFilter::Create() {
  auto ret = std::shared_ptr<CustomFilter>(new CustomFilter());
  gpupixel::GPUPixelContext::GetInstance()->SyncRunWithContext([&] {
    if (ret && !ret->Init()) {
      ret.reset();
    }
  });
  return ret;
}
```

--------------------------------

### Import GPUPixel Header in Objective-C

Source: https://gpupixel.pixpark.net/zh/guide/integrated

Import the main GPUPixel header file in your Objective-C source files. Ensure your Objective-C files have a .mm extension to enable C++ interoperability.

```objective-c
#import <gpupixel/gpupixel.h>
```

--------------------------------

### Manage Frame Buffer

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Create and manage the frame buffer for your input source. Ensure the frame buffer dimensions match the input width and height, and set the output rotation.

```cpp
if (!framebuffer_ || (framebuffer_->GetWidth() != width ||
                      framebuffer_->GetHeight() != height)) {
  framebuffer_ = 
      GPUPixelContext::GetInstance()->GetFramebufferFactory()->CreateFramebuffer(
          width, height);
}
this->SetFramebuffer(framebuffer_, outputRotation);
```

--------------------------------

### 自定义 Sink 实现示例

Source: https://gpupixel.pixpark.net/zh/custom/custom_target

展示了继承 Sink 类并实现自定义渲染逻辑的最小化代码结构。

```cpp
class MyCustomTarget : public Sink {
 public:
  MyCustomTarget() : Sink(1) {
    // 初始化着色器和其他资源
    initShaders();
  }

  ~MyCustomTarget() {
    // 清理资源
    cleanup();
  }

  void Render() override {
    if (!IsReady()) return;
    
    // 设置渲染状态
    setupRenderState();
    
    // 执行自定义渲染逻辑
    Render();
    
    // 处理输出
    processOutput();
  }

 private:
  // 自定义实现细节
};
```

--------------------------------

### Adjust Makeup Filter Parameters

Source: https://gpupixel.pixpark.net/zh/call/beauty_effects

Shows how to adjust the intensity of lipstick and blusher effects using their respective filters. The blend level controls the concentration of the makeup effect.

```cpp
// 设置口红浓度 (0.0-1.0)
lipstick_filter_->SetBlendLevel(value/10);

// 设置腮红浓度 (0.0-1.0)
blusher_filter_->SetBlendLevel(value/10);
```

--------------------------------

### SinkRawData 原始数据输出实现

Source: https://gpupixel.pixpark.net/zh/custom/custom_target

SinkRawData 类支持将渲染结果输出为 RGBA 或 I420 格式的原始数据，并利用 PBO 优化性能。

```cpp
class SinkRawData : public Sink {
 public:
  static std::shared_ptr<SinkRawData> Create();
  void Render() override;

  // 获取输出尺寸与缓冲区（拉取方式，非回调）
  const uint8_t* GetRgbaBuffer();
  const uint8_t* GetI420Buffer();
  int GetWidth() const;
  int GetHeight() const;
};
```

--------------------------------

### Sink 基类接口定义

Source: https://gpupixel.pixpark.net/zh/custom/custom_target

Sink 类是所有输出目标的基类，定义了管理输入帧缓冲区、处理旋转模式和更新机制的核心接口。

```cpp
class Sink {
 public:
  // 构造函数，指定输入数量
  Sink(int inputNumber = 1);
  
  // 设置输入帧缓冲区
  virtual void SetInputFramebuffer(std::shared_ptr<GPUPixelFramebuffer> framebuffer,
                                 RotationMode rotation_mode = NoRotation,
                                 int texIdx = 0);
  
  // 更新处理
  virtual void Render(){};
};
```

--------------------------------

### Implement Custom Filter Rendering Logic in C++

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

The DoRender() method updates shader uniform variables with current filter parameters and then calls the base class implementation to perform the rendering.

```cpp
bool CustomFilter::DoRender(bool updateSinks) {
  // 更新着色器 uniform 变量
  filter_program_->SetUniformValue("parameter", _parameter);
  
  // 调用基类实现
  return Filter::DoRender(updateSinks);
}
```

--------------------------------

### 清理 GPUPixel 资源

Source: https://gpupixel.pixpark.net/zh/call/basic_filters

在完成处理后，通过重置智能指针来释放滤镜链组件占用的资源。

```cpp
// 清理 GPUPixel 资源
// 释放你的滤镜引用
source_raw_input_.reset();
beauty_face_filter_.reset();
target_raw_output_.reset();
// 其他滤镜...
```

--------------------------------

### Process Raw Data Input

Source: https://gpupixel.pixpark.net/zh/call/input_output

Use ProcessData to feed pixel data into the filter chain. Note that only RGBA/BGRA formats are supported; YUV data must be converted first.

```cpp
// 对于 RGBA 数据（需指定 GPUPIXEL_FRAME_TYPE_RGBA 或 GPUPIXEL_FRAME_TYPE_BGRA）
source_raw_input_->ProcessData(pixels, width, height, stride,
                              GPUPIXEL_FRAME_TYPE_RGBA);
```

```cpp
gpuPixelRawInput->ProcessData(pixels, width, height, stride,
                              GPUPIXEL_FRAME_TYPE_RGBA);
```

--------------------------------

### SinkRender 渲染到屏幕实现

Source: https://gpupixel.pixpark.net/zh/custom/custom_target

SinkRender 类用于将图像渲染到屏幕，支持多种填充模式、镜像显示及旋转处理。

```cpp
class SinkRender : public Sink {
 public:
  // 初始化着色器程序和属性位置
  void Init() {
   
  }

  // 实现更新方法，执行实际的渲染
  void Render() override {
     // do render
  }
};
```

--------------------------------

### Adjust Face Reshape Filter Parameters

Source: https://gpupixel.pixpark.net/zh/call/beauty_effects

Illustrates how to control the face slimming and eye enlargement effects using the FaceReshapeFilter. Parameter values are typically between 0.0 and 1.0.

```cpp
// 设置瘦脸程度 (0.0-1.0)
face_reshape_filter_->SetFaceSlimLevel(value/100);

// 设置大眼程度 (0.0-1.0)
face_reshape_filter_->SetEyeZoomLevel(value/50);
```

--------------------------------

### Define Custom Filter Structure in C++

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

This is the basic structure for a custom filter class inheriting from GPUPixel's Filter. It includes methods for creation, initialization, rendering, and parameter setting.

```cpp
class CustomFilter : public Filter {
 public:
  static std::shared_ptr<CustomFilter> Create();
  bool Init();
  virtual bool DoRender(bool updateSinks = true) override;

  // 自定义参数设置器
  void setParameter(float value);

 protected:
  CustomFilter(){};

  // 滤镜参数
  float _parameter;
};
```

--------------------------------

### Process Raw Pixel Data

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Implement `setFrameData` to handle raw pixel data. This involves updating the OpenGL texture with new pixel data, width, height, and stride.

```cpp
void YourCustomInput::setFrameData(const uint8_t* pixels,
                                   int width,
                                   int height,
                                   int stride) {
 
  // Update texture
  glBindTexture(GL_TEXTURE_2D, this->GetFramebuffer()->GetTexture());
  glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0,
               GL_RGBA, GL_UNSIGNED_BYTE, pixels);
}
```

--------------------------------

### Add GPUPixel AAR to Android Project

Source: https://gpupixel.pixpark.net/zh/guide/integrated

Include the gpupixel-release.aar file in your Android project's libs directory and add it as a dependency in your Gradle build file.

```gradle
dependencies {
    implementation files('libs/gpupixel-release.aar')
}
```

--------------------------------

### Control Data Stream Rendering

Source: https://gpupixel.pixpark.net/zh/custom/custom_input

Call `DoRender(true, timestamp)` to process the frame through the filter chain. This method is essential for controlling the data flow and applying filters.

```cpp
Source::DoRender(true, timestamp);
```

--------------------------------

### Define HueFilter Structure in C++

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

This C++ code defines the structure for a HueFilter, inheriting from the base Filter class. It includes methods for creation, initialization, rendering, and a specific setter for hue adjustment.

```cpp
class HueFilter : public Filter {
 public:
  static std::shared_ptr<HueFilter> Create();
  bool Init();
  virtual bool DoRender(bool updateSinks = true) override;

  void setHueAdjustment(float hueAdjustment);

 protected:
  HueFilter(){};
  float _hueAdjustment;
};
```

--------------------------------

### Retrieve Processed Output Data

Source: https://gpupixel.pixpark.net/zh/call/input_output

Extract processed results from the pipeline using SinkRawData after rendering is complete.

```cpp
// 获取 RGBA 结果
const uint8_t* rgba_data = target_raw_output_->GetRgbaBuffer();
int width = target_raw_output_->GetWidth();
int height = target_raw_output_->GetHeight();
size_t rgba_size = width * height * 4;
// 使用 rgba_data 进行处理或显示

// 获取 I420 结果
const uint8_t* i420_data = target_raw_output_->GetI420Buffer();
size_t y_size = width * height;
const uint8_t* uData = i420_data + y_size;
const uint8_t* vData = i420_data + y_size + y_size / 4;
```

--------------------------------

### Define Custom Fragment Shader in GLSL

Source: https://gpupixel.pixpark.net/zh/custom/custom_filter

This GLSL code defines the fragment shader for a custom filter. It includes uniform variables for input texture and custom parameters, and the main function where the image processing logic resides.

```glsl
const std::string kCustomFragmentShader = R"(
    precision highp float;
    uniform sampler2D inputImageTexture;
    uniform float parameter;
    varying highp vec2 textureCoordinate;

    void main() {
      // 在此实现滤镜效果
      vec4 textureColor = texture2D(inputImageTexture, textureCoordinate);
      gl_FragColor = textureColor;
    })"
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.